Require proof of refused reads, fail on missing data, build the BLE server from SERVER_TREE and log when the sleep share is unavailable

Co-authored-by: pre-commit-ci-lite[bot] <117423508+pre-commit-ci-lite[bot]@users.noreply.github.com>
Co-authored-by: Leonardo Rivera <leodrivera@gmail.com>
Co-authored-by: J. Nick Koston <nick@home-assistant.io>

.agents/skills/automations/SKILL.md

@@ -0,0 +1,215 @@
+---
+name: automations
+description: Write ESPHome automation building blocks - triggers (on_* keys), actions and conditions. Use when adding or changing an on_* trigger, an action such as my_component.do_something, a condition, a callback registration method (add_on_*_callback), or code that calls build_callback_automation, build_automation, register_*_action or register_*_condition.
+---
+
+# Automations: triggers, actions and conditions
+
+Automations have three building blocks: **Triggers** (fire when something happens), **Actions** (do
+something), and **Conditions** (check if something is true). The helpers named here are documented in
+their docstrings in `esphome/automation.py`.
+
+## Triggers: callback method (preferred)
+
+Use `build_callback_automation()` for simple triggers. This eliminates the need for a C++ Trigger class by
+using a lightweight pointer-sized forwarder struct registered directly as a callback. No `CONF_TRIGGER_ID`
+in the schema.
+
+**Python:**
+```python
+from esphome import automation
+
+CONFIG_SCHEMA = cv.Schema(
+    {
+        cv.GenerateID(): cv.declare_id(MyComponent),
+        cv.Optional(CONF_ON_STATE): automation.validate_automation({}),
+    }
+).extend(cv.COMPONENT_SCHEMA)
+
+
+async def to_code(config):
+    var = cg.new_Pvariable(config[CONF_ID])
+    await cg.register_component(var, config)
+    for conf in config.get(CONF_ON_STATE, []):
+        await automation.build_callback_automation(
+            var, "add_on_state_callback", [(bool, "x")], conf
+        )
+```
+
+`build_callback_automation` arguments: `parent`, `callback_method` (C++ method name), `args` (template args
+as `[(type, name)]` tuples), `config`, and optional `forwarder` (defaults to `TriggerForwarder<Ts...>`).
+
+For boolean filtering (e.g. `on_press`/`on_release`), use built-in forwarders with `args=[]`:
+```python
+for conf_key, forwarder in (
+    (CONF_ON_PRESS, automation.TriggerOnTrueForwarder),
+    (CONF_ON_RELEASE, automation.TriggerOnFalseForwarder),
+):
+    for conf in config.get(conf_key, []):
+        await automation.build_callback_automation(
+            var, "add_on_state_callback", [], conf, forwarder=forwarder
+        )
+```
+
+When the callback's parameters differ from the automation's arguments, or the trigger should only fire for
+some values, pass `params`, `forward` and `when`; the helper then generates a capture-less lambda (stored
+inline, nothing allocated). Name the parent with `automation.parent_ref(var)` and build `forward` from
+`MockObj` calls, not f-strings of C++:
+```python
+parent = automation.parent_ref(var)
+index = cg.RawExpression("index")
+await automation.build_callback_automation(
+    var,
+    "add_on_state_callback",
+    [(cg.StringRef, "x"), (cg.size_t, "i")],
+    conf,
+    params=[(cg.size_t, "index")],
+    forward=[cg.StringRef(parent.option_at(index)), index],
+)
+```
+`build_parent_callback_automation`, `build_trigger_callback`, `build_callback_automations` and
+`build_trigger_automations` cover the other shapes; see their docstrings in `esphome/automation.py`.
+
+**C++: no trigger class needed.** The callback registration method must be templatized to accept both
+`std::function` and lightweight forwarder structs (which avoid heap allocation):
+```cpp
+class MyComponent : public Component {
+ public:
+  // Must be a template -- accepts both std::function and pointer-sized forwarder structs
+  template<typename F> void add_on_state_callback(F &&callback) {
+    this->state_callback_.add(std::forward<F>(callback));
+  }
+ protected:
+  // Use CallbackManager when callbacks are always registered (e.g. core components)
+  CallbackManager<void(bool)> state_callback_;
+  // Use LazyCallbackManager when callbacks are often not registered -- saves 8 bytes
+  // (nullptr vs empty std::vector) per instance when no callbacks are added
+  // LazyCallbackManager<void(bool)> state_callback_;
+};
+```
+
+Never type a registration method's parameter as `std::function`; that forces a heap allocation for every
+forwarder struct:
+```cpp
+// Bad
+void add_on_state_callback(std::function<void(bool)> &&callback) {
+  this->state_callback_.add(std::move(callback));
+}
+```
+
+| Type | Idle overhead (32-bit) | When to use |
+|------|----------------------|-------------|
+| `CallbackManager<void(Ts...)>` | 12 bytes (empty `std::vector`) | Callbacks are always or almost always registered |
+| `LazyCallbackManager<void(Ts...)>` | 4 bytes (`nullptr`) | Callbacks are often not registered (common case) |
+
+Both live in `esphome/core/helpers.h` and accept `std::function`, lambdas and forwarder structs through
+their templatized `add()` method. `LazyCallbackManager` is a drop-in replacement for `CallbackManager` that defers allocation until the first
+callback is added. Prefer it for entity-level callbacks where most instances have no subscribers.
+
+## Triggers: Trigger class method
+
+Use `build_automation()` with a `Trigger<Ts...>` subclass only when the forwarder needs **mutable state
+beyond a single `Automation*` pointer** (e.g. edge detection tracking previous state, timing logic).
+
+**Python:**
+```python
+TurnOnTrigger = my_ns.class_("TurnOnTrigger", automation.Trigger.template())
+
+CONFIG_SCHEMA = cv.Schema(
+    {
+        cv.Optional(CONF_ON_TURN_ON): automation.validate_automation(
+            {cv.GenerateID(CONF_TRIGGER_ID): cv.declare_id(TurnOnTrigger)}
+        ),
+    }
+)
+
+
+async def to_code(config):
+    for conf in config.get(CONF_ON_TURN_ON, []):
+        trigger = cg.new_Pvariable(conf[CONF_TRIGGER_ID], var)
+        await automation.build_automation(trigger, [], conf)
+```
+
+**C++:**
+```cpp
+class TurnOnTrigger : public Trigger<> {
+ public:
+  explicit TurnOnTrigger(MyComponent *parent) : last_on_{false} {
+    parent->add_on_state_callback([this](bool state) {
+      if (state && !this->last_on_)
+        this->trigger();
+      this->last_on_ = state;
+    });
+  }
+ protected:
+  bool last_on_;
+};
+```
+
+## Actions
+
+```cpp
+template<typename... Ts> class MyAction : public Action<Ts...> {
+ public:
+  explicit MyAction(MyComponent *parent) : parent_(parent) {}
+  void play(const Ts &...) override { this->parent_->do_something(); }
+ protected:
+  MyComponent *parent_;
+};
+```
+Register it without writing a builder:
+```python
+automation.register_simple_action(
+    "my_component.do_something", MyAction, schema, synchronous=True
+)
+```
+The constructor receives the object named by `config[CONF_ID]`. Use `register_bare_action` for a
+no-argument constructor, `register_parented_action` for a class deriving from `Parented<T>`, and the
+`@automation.register_action(...)` decorator only when the builder must also set fields.
+
+Use `synchronous=True` for actions that run to completion inside `play()` without deferring. Use
+`synchronous=False` if the action may suspend/defer execution (e.g. `delay`, `wait_until`, `script.wait`)
+or store trigger arguments for later use.
+
+**Actions that only forward templatable values to their parent need no C++ class.** Register them with
+`register_apply_action`; do not write a `TEMPLATABLE_VALUE` class or a builder for this shape.
+```python
+automation.register_apply_action(
+    "my_component.set_gains",
+    schema,
+    automation.ApplyField(CONF_KP, "set_kp", cg.float_),
+    automation.ApplyField(CONF_KI, "set_ki", cg.float_),
+)
+```
+The `ApplyField`, `ApplyCall` and `register_apply_action` docstrings in `esphome/automation.py` cover the
+rest; `cover.control` and `cover.template.publish` are in-tree examples. An action whose input is a byte
+payload passes `automation.ApplyBytes(CONF_DATA, "send({})")` in the same call; a constant payload stays in
+flash and a lambda's vector is forwarded as data and size. `TEMPLATABLE_VALUE` with `cg.templatable` stays
+for actions whose `play()` has real logic beyond forwarding values.
+
+## Conditions
+
+```cpp
+template<typename... Ts> class MyCondition : public Condition<Ts...> {
+ public:
+  explicit MyCondition(MyComponent *parent) : parent_(parent) {}
+  bool check(const Ts &...) override { return this->parent_->is_active(); }
+ protected:
+  MyComponent *parent_;
+};
+```
+Register with `automation.register_simple_condition("my_component.is_active", MyCondition, schema)`;
+`register_bare_condition`, `register_parented_condition` and the decorator follow the action rules.
+
+**Conditions that only test their parent need no C++ class either.** Register them with
+`register_apply_condition`; the expression is applied to the parent, and an `ApplyCall` compares against
+config values.
+```python
+automation.register_apply_condition("my_component.is_active", schema, "is_active()")
+automation.register_apply_condition(
+    "my_component.state_is",
+    schema,
+    automation.ApplyCall("state == {}", ((CONF_STATE, cg.bool_),)),
+)
+```
+`cover.is_open`, `rtttl.is_playing` and `component.is_idle` are in-tree examples.

.agents/skills/breaking-changes/SKILL.md

@@ -0,0 +1,68 @@
+---
+name: breaking-changes
+description: Decide whether an ESPHome change breaks public API and handle it - what counts as public C++ and Python API, the deprecation window, ESPDEPRECATED for C++, cv.rename_key and deprecation warnings for config keys, and the migration notes a PR needs. Use when removing, renaming or changing the signature of a config option, a public C++ member or a core Python helper.
+---
+
+# Public API and breaking changes
+
+Full policy: https://developers.esphome.io/contributing/code/#public-api-and-breaking-changes
+
+## What is public
+
+- **C++ in components:** only features documented at [esphome.io](https://esphome.io) are public API.
+  Undocumented `public` members are internal.
+- **C++ core and base classes** (`esphome/core/`, `Component`, `Sensor`, etc.): all `public` members are
+  public API.
+- **Components with global accessors** (`global_api_server`, etc.): all `public` members are public API
+  (except config setters).
+- **Python:** all documented configuration options are public API. Python code in `esphome/core/` that
+  existing core components use is stable API. Other Python code is internal unless explicitly documented
+  for external component use.
+- Changing the signature of a lambda that code generation injects is not a breaking change.
+
+## Policy
+
+- Aim for a **6-month deprecation window** when possible.
+- Clean breaks are allowed for signature changes, deep refactorings and resource constraints.
+- The migration path goes in the PR description (it generates the release notes).
+- A blog post is required for core/base class changes or significant architectural changes.
+
+Checklist:
+- [ ] Clear justification (RAM/flash savings, architectural improvement)
+- [ ] Explored non-breaking alternatives
+- [ ] Added deprecation warnings if possible (`ESPDEPRECATED` for C++)
+- [ ] Documented migration path in the PR description with before/after examples
+- [ ] Updated all internal usage and esphome.io
+- [ ] Tested backward compatibility during the deprecation period
+
+## Deprecating C++
+
+```cpp
+// Remove before 2026.6.0
+ESPDEPRECATED("Use new_method() instead. Removed in 2026.6.0", "2025.12.0")
+void old_method() { this->new_method(); }
+```
+
+## Deprecating config keys
+
+For a renamed key, use the shared `cv.rename_key` validator with `removed_in` (and `component` for
+context); it warns and migrates automatically:
+```python
+CONFIG_SCHEMA = cv.All(
+    cv.rename_key(
+        CONF_OLD_KEY, CONF_NEW_KEY, removed_in="2026.6.0", component="my_component"
+    ),
+    cv.Schema({...}),
+)
+```
+For other deprecations, warn manually during validation:
+```python
+# Remove before 2026.6.0
+if CONF_OLD_KEY in config:
+    _LOGGER.warning(
+        f"'{CONF_OLD_KEY}' deprecated, use '{CONF_NEW_KEY}'. Removed in 2026.6.0"
+    )
+    config[CONF_NEW_KEY] = config.pop(CONF_OLD_KEY)  # Auto-migrate
+```
+Cover the old syntax with a `validate.*.yaml` test (see the `component-tests` skill), and keep the language
+schema correct (see the `config-validation` skill).

.agents/skills/code-review/SKILL.md

@@ -9,11 +9,13 @@ ESPHome parses YAML into C++ firmware for memory-constrained microcontrollers
 (ESP32, ESP8266, RP2040, LibreTiny). Review changes with that in mind: RAM and
 flash are scarce, and code runs unattended for months.
 
-`AGENTS.md` in the repository root is the full contributor guide and the
-authority when it disagrees with this summary. The developer documentation at
-https://developers.esphome.io explains the component lifecycle and the reasoning
-behind these rules. This skill lists the concrete things worth flagging in a
-review; read `AGENTS.md` for the detail behind any item.
+`AGENTS.md` in the repository root is the contributor guide, and the skills it
+lists (`component-codegen`, `config-validation`, `automations`,
+`embedded-memory`, `component-tests`, `breaking-changes`) hold the detail. They
+are the authority when they disagree with this summary. The developer
+documentation at https://developers.esphome.io explains the component lifecycle
+and the reasoning behind these rules. This skill lists the concrete things worth
+flagging in a review; read the matching skill for the detail behind any item.
 
 Only raise findings that the diff actually introduces or changes. Do not ask for
 drive-by cleanup of pre-existing code the PR did not touch.
@@ -76,6 +78,10 @@ performance nit, because it fragments a small shared heap. Flag:
   `if (blah := config.get(CONF_BLAH)) is not None:`.
 - Reuse existing validators from `config_validation.py` (`cv.rename_key`,
   `cv.has_exactly_one_key`, etc.) via `cv.All(...)` instead of hand-rolling.
+- A new or changed validator function that wraps, chooses or builds a schema
+  must be decorated with `@schema_extractor(...)` and handle `SCHEMA_EXTRACT`,
+  or `script/build_language_schema.py` silently drops the keys behind it (see
+  the `config-validation` skill).
 - `esphome/const.py` is frozen: no new `CONF_` constants there. Define them in
   the component's own `.py`, or in `esphome/components/const/__init__.py` when
   shared. The same constant defined in three or more component files fails CI.

.agents/skills/component-codegen/SKILL.md

@@ -0,0 +1,241 @@
+---
+name: component-codegen
+description: Write the Python side of an ESPHome component - __init__.py and platform files (sensor.py, switch.py, ...), DOMAIN and other metadata, to_code, hub child entities with sub_sensors and friends, pins and referenced components, hub-forwarding entities (ApplyButton, ApplySwitch), uart/i2c/spi constructor helpers, and CORE.data state. Use when creating a component or changing its code generation.
+---
+
+# Component code generation
+
+Read https://developers.esphome.io/architecture/components/ for the component lifecycle and how a
+component is registered. Schema rules and schema extraction are in the `config-validation` skill;
+automations (`on_*`, actions, conditions) in the `automations` skill.
+
+## Files
+
+```
+components/[component_name]/
+├── __init__.py          # Component configuration schema and code generation
+├── [component].h        # C++ header file (if needed)
+├── [component].cpp      # C++ implementation (if needed)
+└── [platform]/          # Platform-specific implementations
+    ├── __init__.py      # Platform-specific configuration
+    ├── [platform].h     # Platform C++ header
+    └── [platform].cpp   # Platform C++ implementation
+```
+
+## Metadata
+
+- `DOMAIN`: required in every component's `__init__.py`, set to the component's own name as a plain string
+  (e.g. `DOMAIN = "my_component"`). Other code refers to the component through it, e.g.
+  `from esphome.components.my_component import DOMAIN`. Keep it in alphabetical order with the other
+  metadata constants. CI (`lint_component_domain`) fails without it.
+- `DEPENDENCIES`: list of required components.
+- `AUTO_LOAD`: components to automatically load. Not a way to put entities under the component's own key;
+  see "Entities of a hub" below.
+- `CONFLICTS_WITH`: incompatible components.
+- `CODEOWNERS`: GitHub usernames responsible for maintenance.
+- `MULTI_CONF`: whether multiple instances are allowed.
+
+## Basic pattern
+
+```python
+import esphome.codegen as cg
+import esphome.config_validation as cv
+from esphome.const import CONF_KEY, CONF_ID
+
+DOMAIN = "my_component"
+CONF_PARAM = "param"  # A constant that does not yet exist in esphome/const.py
+
+my_component_ns = cg.esphome_ns.namespace("my_component")
+MyComponent = my_component_ns.class_("MyComponent", cg.Component)
+
+CONFIG_SCHEMA = cv.Schema(
+    {
+        cv.GenerateID(): cv.declare_id(MyComponent),
+        cv.Required(CONF_KEY): cv.string,
+        cv.Optional(CONF_PARAM, default=42): cv.int_,
+    }
+).extend(cv.COMPONENT_SCHEMA)
+
+
+async def to_code(config):
+    var = cg.new_Pvariable(config[CONF_ID])
+    await cg.register_component(var, config)
+    cg.add(var.set_key(config[CONF_KEY]))
+    cg.add(var.set_param(config[CONF_PARAM]))
+```
+
+```cpp
+namespace esphome::my_component {
+
+class MyComponent : public Component {
+ public:
+  void setup() override;
+  void loop() override;
+  void dump_config() override;
+
+  void set_key(const char *key) { this->key_ = StringRef(key); }
+  void set_param(int param) { this->param_ = param; }
+
+ protected:
+  StringRef key_;
+  int param_{0};
+};
+
+}  // namespace esphome::my_component
+```
+
+When the setter takes the value unchanged, use `cg.add_optional(var.set_blah, config.get(CONF_BLAH))`; it
+is not awaited, and it skips only `None`, so `0` and `False` are still set. Required, invariant
+dependencies are constructor arguments to `cg.new_Pvariable()`, not setters.
+
+## Entity platforms
+
+**Sensor:**
+```python
+from esphome.components import sensor
+
+CONFIG_SCHEMA = sensor.sensor_schema(MySensor).extend(
+    cv.polling_component_schema("60s")
+)
+
+
+async def to_code(config):
+    var = await sensor.new_sensor(config)
+    await cg.register_component(var, config)
+```
+
+**Binary sensor:**
+```python
+from esphome.components import binary_sensor
+
+CONFIG_SCHEMA = binary_sensor.binary_sensor_schema().extend({...})
+
+
+async def to_code(config):
+    var = await binary_sensor.new_binary_sensor(config)
+```
+
+**Switch:**
+```python
+from esphome.components import switch
+
+CONFIG_SCHEMA = switch.switch_schema().extend({...})
+
+
+async def to_code(config):
+    var = await switch.new_switch(config)
+```
+
+## Entities of a hub
+
+A new component's entities go in platform files (`sensor.py`, `binary_sensor.py`, ...) that take the hub's
+id, never under the hub's own key with their platforms added to `AUTO_LOAD` to make that work. Auto loading
+a platform because the C++ uses its type is fine. See
+https://developers.esphome.io/architecture/components/#exposing-multiple-entities.
+```yaml
+# Bad - entities under the hub's key
+my_hub:
+  connected:
+    name: Connected
+
+# Good - the hub, then one platform entry that holds its entities
+my_hub:
+  id: hub1
+binary_sensor:
+  - platform: my_hub
+    my_hub_id: hub1  # optional with a single hub
+    connected:
+      name: Connected
+```
+
+**Optional child entities of a hub:** bind the config once with `sensor.sub_sensors(config)` (or
+`sub_binary_sensors`, `sub_text_sensors`, `sub_buttons`, `sub_switches`, `sub_numbers`, `sub_selects` in
+their domains), adding `parent=hub` for entities that derive from `Parented<T>`, then make one call per
+key, even when there is only one. A call creates the entity only when its key is configured, passes it to
+the setter and returns it (or `None`); extra arguments such as `min_value` or `options` go on the call.
+Always name the setter explicitly on the object that owns it, never with `getattr` and an f-string, and
+keep that variable short (`var` for the component itself, `hub` for one fetched with `cg.get_variable`) so
+the calls fit on one line. Loops whose setter also takes an index, such as `set_gate_threshold(x, n)`, stay
+as they are.
+```python
+# my_hub/sensor.py
+async def to_code(config):
+    hub = await cg.get_variable(config[CONF_MY_HUB_ID])
+    sensors = sensor.sub_sensors(config)
+    await sensors(CONF_TEMPERATURE, hub.set_temperature_sensor)
+    await sensors(CONF_HUMIDITY, hub.set_humidity_sensor)
+
+
+# my_hub/button.py
+async def to_code(config):
+    hub = await cg.get_variable(config[CONF_MY_HUB_ID])
+    buttons = button.sub_buttons(config, parent=hub)
+    await buttons(CONF_RESTART, hub.set_restart_button)
+```
+
+**Hub forwarding entities need no C++ class:** a button, switch, number or select that only calls its hub
+uses `button.ApplyButton` (`switch.ApplySwitch`, `number.ApplyNumber`, `select.ApplySelect`) in its schema
+and `apply=` on the `sub_*` call. The body is a call on the hub with `entity` and `state`, `value` or
+`index`/`option` in scope; on a switch, number or select `publish="before"` or `"after"` publishes the
+state, and `publish="on_success"` publishes only when the hub call returns true (a write the device can
+refuse). `publish="momentary"` is only for legacy action switches (restart, factory reset): it publishes off
+and runs the body only when turned on. Never use it for a new entity; an action is a button.
+```python
+switches = switch.sub_switches(config, parent=hub)
+await switches(CONF_BLUETOOTH, None, apply="set_bluetooth(state)", publish="after")
+```
+Restore an owned or standalone switch in `setup()` with `sw->restore()`, or `restore(default)` to still
+drive a state when the restore mode is disabled. A hub that applies its switches' restored states itself
+reads `sw->get_initial_write_state()`, which already applies `inverted`.
+
+**Entities that each watch a register field:** one class taking register, shift, mask and match from a
+Python tuple table, in typed `StaticVector`s sized by `cg.slot_counter`; not a template per bit, not a
+virtual listener base (a second vptr per entity).
+
+## Other helpers
+
+- **Pins and referenced components:** set them with `await cg.add_pin(var.set_cs_pin, config[CONF_CS_PIN])`
+  and `await cg.add_variable(var.set_bus, config[CONF_BUS_ID])`; both skip a `None` value, so pass
+  `config.get(KEY)` for an optional key, and return what they created. Use `cg.register_parented` for
+  `set_parent`.
+- `uart.new_uart_component(config, *ctor_args)` (and `i2c.new_i2c_component`, `spi.new_spi_component`,
+  `modbus.new_modbus_client_component`, `ble_client.new_ble_node_component`) constructs, registers and
+  attaches in one call.
+- `display.add_writer(var, config)` sets the lambda.
+- `MyEnum.options("on", prefix="MY_ENUM_")` builds an enum dict.
+- `button.restart_button_schema(cls)` (and `factory_reset_`, `query_`) are button presets.
+- `fv.get_declaring_config(id)` finds a hub's config in final validation.
+
+## State during code generation
+
+Use `CORE.data` for component state that needs to persist during configuration generation, namespaced
+under the component's `DOMAIN` and held in a `@dataclass`. Avoid module-level mutable globals: they
+persist between compilation runs if the host process (e.g. device-builder) doesn't fork/exec, whereas
+`CORE.data` is cleared between runs.
+
+```python
+from dataclasses import dataclass, field
+from esphome.core import CORE
+
+DOMAIN = "my_component"
+
+
+@dataclass
+class MyComponentData:
+    feature_enabled: bool = False
+    item_count: int = 0
+    items: list[str] = field(default_factory=list)
+
+
+def _get_data() -> MyComponentData:
+    if DOMAIN not in CORE.data:
+        CORE.data[DOMAIN] = MyComponentData()
+    return CORE.data[DOMAIN]
+
+
+def request_feature() -> None:
+    _get_data().feature_enabled = True
+```
+
+Don't use flat keys such as `CORE.data["my_component_feature"] = True`. Some components use `TypedDict`
+instead of a dataclass; both are acceptable.

.agents/skills/component-tests/SKILL.md

@@ -0,0 +1,101 @@
+---
+name: component-tests
+description: Write and run tests for ESPHome components - YAML build tests (test.*.yaml, validate.*.yaml, common.yaml) under tests/components, shared bus packages and CI test grouping, C++ gtest unit tests, and host integration tests including the uart_mock fixture. Use when adding or changing files under tests/, or when a code change needs test coverage.
+---
+
+# Component tests
+
+## C++ unit tests
+
+gtests live in `tests/components/<component>/*.cpp` and run with `script/cpp_unit_test.py <component>`; see
+`tests/components/README.md`. Set the loop clock in a test with `LoopBlockingGuard{nullptr, nullptr,
+now_ms}`; the runner resets it before every test.
+
+## Integration tests
+
+`tests/integration/` runs real firmware on the host. Serial components prove timeouts and recovery with the
+`uart_mock` fixture (`tests/integration/test_uart_mock_*.py`). Its `responses` match on content every time
+they see it; to answer a repeated request differently, use an `on_tx` lambda with a static counter and
+`inject_to_rx_buffer` (see `uart_mock_ld2412_requery_after_giveup.yaml`).
+
+## YAML build tests
+
+```
+tests/
+├── test_build_components/
+│   └── common/          # Shared bus packages (uart, i2c, spi, etc.)
+│       ├── uart/        # UART at default baud rate
+│       ├── uart_115200/ # UART at 115200 baud
+│       ├── i2c/         # I2C bus
+│       └── spi/         # SPI bus
+└── components/[component]/
+    ├── common.yaml          # Component-only config (no bus definitions)
+    ├── test.esp32-idf.yaml      # config + compile
+    ├── test.esp8266-ard.yaml    # config + compile
+    ├── test-variant.esp32-idf.yaml  # variant test, config + compile
+    ├── validate.esp32-idf.yaml      # config-only (never compiled)
+    └── validate-legacy.esp32-idf.yaml  # config-only variant
+```
+Run them with `script/test_build_components`. Use `-c <component>` to test specific components and
+`-t <target>` for specific platforms.
+
+- **Keep build-test variants few:** one `test.<platform>.yaml` per platform the code actually differs on.
+  A variant per board for platform-identical code only adds CI time.
+- **Config-only test files (`validate.*.yaml`):** use this prefix when a YAML file only needs to exercise
+  schema/validation paths and does not need to be compiled. CI runs `validate.*.yaml` files with
+  `esphome config` only and skips them during compile. The grammar mirrors `test.*.yaml`:
+  - `validate.<platform>.yaml`: base config-only test
+  - `validate-<variant>.<platform>.yaml`: config-only variant
+
+  Use this for deprecated-syntax migration tests, schema edge cases, or platform-specific validation
+  branches where building firmware adds no signal. A component may have any mix of `test.*.yaml` and
+  `validate.*.yaml` files. Validate files never participate in bus grouping; each one runs as its own
+  `esphome config` invocation.
+
+  When a PR's only edits to a component are `validate.*.yaml` files (no source changes, no `test.*.yaml`
+  changes, and the component isn't pulled in as a dependency of another changed component), CI skips the
+  compile stage for that component and only runs config validation. This is decided in
+  `script/determine-jobs.py` via `_component_change_is_validate_only` and surfaced as the
+  `validate_only_components` output that the `test-build-components-split` job consumes.
+
+## Grouping with packages
+
+Components that use shared bus packages can be grouped together in CI to reduce build count. **Never define
+buses (uart, i2c, spi, modbus) directly in test YAML files**; always use packages from
+`test_build_components/common/`.
+
+All includes in test files must go through dict-style `packages:` so that batch grouping works; the
+grouping scripts only understand dict-style packages. Never use list-style packages
+(`packages: [- !include ...]`) or top-level merge keys (`<<: !include common.yaml`). Bus packages are keyed
+by the bus name; the component's `common.yaml` is keyed by the component name (e.g.
+`cst328: !include common.yaml`):
+```yaml
+# test.esp32-idf.yaml - everything included via named packages
+packages:
+  uart: !include ../../test_build_components/common/uart_115200/esp32-idf.yaml
+  my_component: !include common.yaml
+```
+```yaml
+# common.yaml - component config only, NO bus definitions
+my_component:
+  id: my_instance
+
+sensor:
+  - platform: my_component
+    name: My Sensor
+```
+Components that define buses directly are flagged as "NEEDS MIGRATION" and cannot be grouped, increasing
+CI build time. `!extend` and `!remove` in a `test.*.yaml` also make the component ungroupable; add a
+standalone entity with its own id instead.
+
+A test that needs a shared component in a particular state (`esp32_ble` security settings, an option that
+PSRAM rejects) cannot be grouped: the merged block wins and the whole batch fails. Add the component to
+`ISOLATED_COMPONENTS` in `script/analyze_component_buses.py`, or move the option to a `test-<variant>` file.
+
+To verify that all components can be tested together without ID conflicts or configuration issues:
+```bash
+./script/test_component_grouping.py -e config --all
+```
+This tests all components in a single build to catch conflicts that might not appear when testing
+components individually. Use `-e config` for fast configuration validation, or `-e compile` for full
+compilation testing.

.agents/skills/config-validation/SKILL.md

@@ -0,0 +1,124 @@
+---
+name: config-validation
+description: Write or change ESPHome configuration validation - CONFIG_SCHEMA, cv.Schema, cv.typed_schema, custom validator functions, config keys and CONF_ constants, renamed or deprecated keys, and the language schema that script/build_language_schema.py extracts from them. Use whenever a change touches a component's schema or a validator in config_validation.py, automation.py or a component's .py files.
+---
+
+# Configuration validation
+
+ESPHome validates YAML with `voluptuous` schemas built from `esphome/config_validation.py` (`cv`). The
+same schemas are also walked by `script/build_language_schema.py`, which produces the JSON language schema
+that editors and the dashboard use for completion and checking. A validation change is only finished when
+both still work.
+
+## Writing schemas
+
+- **Type hints:** validators are typed like any other function: `def validate_x(config: ConfigType) ->
+  ConfigType:`, with `ConfigType` imported from `esphome.types`.
+- **One schema per model:** when a component's options depend on a `model` key, use
+  `cv.typed_schema({...}, key=CONF_MODEL, default_type=...)` with each model's own keys and defaults, not
+  one schema plus a validator that rejects keys and fills defaults afterwards.
+- **Reuse existing validators:** before writing a custom validator, check for an existing one in
+  `config_validation.py` and compose it in `cv.All(...)` rather than duplicating logic across components.
+  For example, rename a config key with `cv.rename_key(CONF_OLD, CONF_NEW, removed_in="2026.6.0")`, and
+  reject mutually-exclusive keys with `cv.has_at_most_one_key(...)` / `cv.has_exactly_one_key(...)`. See
+  how `api` composes `cv.has_exactly_one_key` + `cv.rename_key`.
+- **Common validators:** `cv.int_`, `cv.float_`, `cv.string`, `cv.boolean`, `cv.int_range(min=0,
+  max=100)`, `cv.positive_int`, `cv.percentage`.
+- **Complex validation:** `cv.All(cv.string, cv.Length(min=1, max=50))`, `cv.Any(cv.int_, cv.string)`.
+- **Platform-specific:** `cv.only_on(["esp32", "esp8266"])`, `esp32.only_on_variant(...)`,
+  `cv.only_on_esp32`, `cv.only_on_esp8266`, `cv.only_on_rp2040`.
+- **Framework-specific:** `cv.only_with_framework(...)`, `cv.only_with_arduino`. On the ESP32, Arduino is
+  built as a component of ESP-IDF, so an ESP-IDF API, sdkconfig option or version floor is there on both
+  frameworks: gate on `esp32.require_idf_version(cv.Version(5, 5, 3), "psram_dma")`, which checks the
+  platform and the ESP-IDF under either framework, as `spi`, `usb_host` and `wifi` do. Reserve
+  `cv.only_with_framework("esp-idf")` for a build that has been tried on Arduino and fails there, and write
+  the actual failure next to it (`sendspin`: arduino-esp32 below ESP-IDF 6.0 bundles a second libsodium,
+  and from 6.0 sendspin-cpp's enumerators collide with `Arduino.h` macros such as `DISPLAY`).
+- **Schema extensions:**
+  ```python
+  CONFIG_SCHEMA = cv.Schema({ ... })
+   .extend(cv.COMPONENT_SCHEMA)
+   .extend(uart.UART_DEVICE_SCHEMA)
+   .extend(i2c.i2c_device_schema(0x48))
+   .extend(spi.spi_device_schema(cs_pin_required=True))
+  ```
+- **Constants:** `esphome/const.py` is frozen; do not add new `CONF_` constants there. Define a
+  component-local constant in the component's own `.py`; for a constant shared by multiple components, add
+  it to `esphome/components/const/__init__.py`. CI (`lint_constants_usage`) fails if the same constant is
+  defined in three or more component files. Constants used in core files (those not under
+  `esphome/components`) may be added to `esphome/const.py` but will require adjustment to the CI check.
+- **Renamed or removed keys:** see the `breaking-changes` skill for `cv.rename_key` and manual
+  deprecation warnings.
+- **Config-only tests:** schema edge cases and deprecated-syntax migrations go in `validate.*.yaml` test
+  files (see the `component-tests` skill).
+
+## Schema extraction
+
+`script/build_language_schema.py` sets `esphome.schema_extractors.EnableSchemaExtraction = True` before it
+imports any component, then walks every `CONFIG_SCHEMA` (and platform schema) recursively. It understands
+`cv.Schema`, `cv.All`, `cv.Any`, dicts of `cv.Required` / `cv.Optional` keys, `.extend()`, `cv.ensure_list`,
+`cv.typed_schema`, registries and the known `cv` validators. **A plain function it cannot see into**: a
+custom validator that builds or picks a schema at validation time hides every key behind it, and the
+dumper silently emits nothing for them. CI only runs the script with `--check`, which does not compare the
+output, so a lost schema is not caught.
+
+When a validator function wraps, chooses or builds a schema (rather than checking a single value), make it
+visible to the dumper:
+
+1. Decorate it with `@schema_extractor("<type>")` from `esphome.schema_extractors`.
+2. At the top of the function, return the data for that type when it is called with `SCHEMA_EXTRACT`
+   instead of a config value. Compare with `is` (`value is SCHEMA_EXTRACT`); older code uses `==`.
+
+```python
+from esphome.schema_extractors import SCHEMA_EXTRACT, schema_extractor
+
+
+@schema_extractor("schema")
+def _outgoing_connection_schema(config: ConfigType | None) -> ConfigType:
+    if config is SCHEMA_EXTRACT:
+        return _OUTGOING_CONNECTION_SCHEMA
+    if config is None:
+        config = {}
+    return _OUTGOING_CONNECTION_SCHEMA(config)
+```
+
+The type names the dumper accepts, and what `SCHEMA_EXTRACT` must return for each, are in the
+`hidden_schemas` branch of `convert()` in `script/build_language_schema.py`. The common ones:
+
+| Type | Return for `SCHEMA_EXTRACT` | Example |
+|------|-----------------------------|---------|
+| `"schema"` | the schema (or a representative one) to walk | `api` `_outgoing_connection_schema`, `restore_state_schema` in `esphome/core/entity_helpers.py`, `model_schema_extractor` in `esphome/components/mipi/__init__.py` |
+| `"enum"` | the dict of options | `sensor_entity_category` in `esphome/components/sensor/__init__.py` |
+| `"one_of"` | the list of options | `cv.one_of` |
+| `"maybe"` | `(validator, key)` for a scalar shorthand | `cv.maybe_simple_value`, `maybe_conf` in `esphome/automation.py` |
+| `"templatable"` | the inner validator | `cv.templatable` |
+| `"automation"` | the automation schema | `automation.validate_automation` |
+
+Points to watch:
+
+- The dumper matches the function object by `repr`, so the decorated function must be the exact object
+  placed in the schema. In a factory that returns a new closure per call, decorate the inner function (as
+  `restore_state_schema` and `cv.maybe_simple_value` do), not the factory.
+- The decorator only records anything when extraction is enabled, and that flag is set before imports, so
+  the decorator must run at import time (module level or inside a factory called at module level).
+- A schema chosen per config value (for example by `model`) has no single answer: return a representative
+  schema, as `model_schema_extractor` does.
+- Plain value checks (`cv.All(cv.string, my_check)`) need nothing: the dumper still sees the surrounding
+  types and keys.
+- `cv.Schema`, `cv.typed_schema`, `cv.ensure_list` and the `cv` helpers are already handled; prefer
+  composing them over a custom function, which also avoids the extractor work.
+
+### Checking the output
+
+Build the language schema into a scratch directory and look at the component's JSON (it takes about ten
+seconds):
+
+```bash
+python3 script/build_language_schema.py --output-path /tmp/schema
+python3 -c "import json; print(json.dumps(json.load(open('/tmp/schema/my_component.json')), indent=1))" | less
+```
+
+Every key a user can write should appear under `config_vars` with a sensible `type`; compare against the
+same output from `dev` when changing an existing schema. If the script prints `Unknown extracted schema
+type` or raises, the decorator type or the value returned for `SCHEMA_EXTRACT` is wrong. Unit tests for
+the dumper live in `tests/script/test_build_language_schema.py`.

.agents/skills/embedded-memory/SKILL.md

@@ -0,0 +1,169 @@
+---
+name: embedded-memory
+description: Choose ESPHome C++ containers and storage that avoid heap use and fragmentation on microcontrollers - std::array, StaticVector with cg.slot_counter, FixedVector, std::unique_ptr byte buffers, StringRef, RAMAllocator, and what to avoid (std::vector growth, std::map, std::deque, std::string, new (std::nothrow)). Use when adding or changing C++ members, containers, buffers, strings or registration lists, or when reducing RAM or flash use.
+---
+
+# Memory on embedded targets
+
+ESP devices run for months with small heaps shared between Wi-Fi, BLE, LWIP, and application code. Over
+time, repeated allocations of different sizes fragment the heap. Failures happen when the largest
+contiguous block shrinks, even if total free heap is still large. We have seen field crashes caused by
+this.
+
+**Heap allocation after `setup()` should be avoided unless absolutely unavoidable.** Every
+allocation/deallocation cycle contributes to fragmentation. ESPHome treats runtime heap allocation as a
+long-term reliability bug, not a performance issue. Helpers that hide allocation (`std::string`,
+`std::to_string`, string-returning helpers) are being deprecated and replaced with buffer and view based
+APIs.
+
+## Container guidelines
+
+1. **Compile-time-known sizes:** use `std::array` instead of `std::vector` when size is known at compile
+   time.
+   ```cpp
+   // Bad - generates STL realloc code
+   std::vector<int> values;
+
+   // Good - no dynamic allocation
+   std::array<int, MAX_VALUES> values;
+   ```
+   Use `cg.add_define("MAX_VALUES", count)` to set the size from Python configuration.
+
+   **For byte buffers:** avoid `std::vector<uint8_t>` unless the buffer needs to grow. Use
+   `std::unique_ptr<uint8_t[]>` instead. It does **not** provide bounds checking or iterator support like
+   `std::vector<uint8_t>`; use it only when you do not need these features.
+   ```cpp
+   // Bad - STL overhead for simple byte buffer
+   std::vector<uint8_t> buffer;
+   buffer.resize(256);
+
+   // Good - minimal overhead, single allocation
+   std::unique_ptr<uint8_t[]> buffer = std::make_unique<uint8_t[]>(256);
+   // Or if size is constant:
+   std::array<uint8_t, 256> buffer;
+   ```
+
+2. **Compile-time-known fixed sizes with vector-like API:** use `StaticVector` from
+   `esphome/core/helpers.h` for compile-time fixed size with a `push_back()` interface (no dynamic
+   allocation).
+   ```cpp
+   // Bad - generates STL realloc code (_M_realloc_insert)
+   std::vector<ServiceRecord> services;
+   services.reserve(5);  // Still includes reallocation machinery
+
+   // Good - compile-time fixed size, no dynamic allocation
+   StaticVector<ServiceRecord, MAX_SERVICES> services;
+   services.push_back(record1);
+   ```
+   Use `cg.add_define("MAX_SERVICES", count)` to set the size from Python configuration.
+
+   **Listener and child-entity registration lists are the most common case, and the most commonly
+   missed.** A `register_*()` method called once per child at code generation time has a count that is
+   known at compile time, so it should never be a `std::vector`. Use `cg.slot_counter()`: it returns a
+   function that each consumer calls once per slot it will occupy, and after every `to_code` has run it
+   emits the define with the final count. When nothing registers, no define is emitted and the storage plus
+   its registration method compile out entirely.
+   ```python
+   # hub component's __init__.py
+   _request_listener_slot = cg.slot_counter("MY_COMPONENT_LISTENER_COUNT")
+
+
+   async def register_listener(hub: MockObj, var: MockObj) -> None:
+       _request_listener_slot()
+       cg.add(hub.register_listener(var))
+   ```
+   When several instances each own a list declared at the same size (one per hub of a `MULTI_CONF`
+   component), pass the owning object as the key, `_request_listener_slot(str(hub))`; the define is then
+   the largest count any one key requested instead of the total.
+   ```cpp
+   #ifdef MY_COMPONENT_LISTENER_COUNT
+     void register_listener(MyComponentListener *listener);
+   #endif
+    protected:
+   #ifdef MY_COMPONENT_LISTENER_COUNT
+     StaticVector<MyComponentListener *, MY_COMPONENT_LISTENER_COUNT> listeners_;
+   #endif
+   ```
+   Request slots from `to_code`, not from a job that runs after `CoroPriority.FINAL`; a late request
+   raises rather than silently undercounting.
+
+3. **Runtime-known sizes:** use `FixedVector` from `esphome/core/helpers.h` when the size is only known at
+   runtime initialization.
+   ```cpp
+   // Bad - generates STL realloc code (_M_realloc_insert)
+   std::vector<TxtRecord> txt_records;
+   txt_records.reserve(5);  // Still includes reallocation machinery
+
+   // Good - runtime size, single allocation, no reallocation machinery
+   FixedVector<TxtRecord> txt_records;
+   txt_records.init(record_count);  // Initialize with exact size at runtime
+   ```
+   This eliminates `_M_realloc_insert` and `_M_default_append` template instantiations (200-500 bytes per
+   instance), needs no upper bound, and is compatible with protobuf code generation when using
+   `[(fixed_vector) = true]`.
+
+4. **Small datasets (1-16 elements):** use `std::vector` or `std::array` with simple structs instead of
+   `std::map`/`std::set`/`std::unordered_map`.
+   ```cpp
+   // Bad - 2KB+ overhead for red-black tree/hash table
+   std::map<std::string, int> small_lookup;
+   std::unordered_map<int, std::string> tiny_map;
+
+   // Good - simple struct with linear search (std::vector is fine)
+   struct LookupEntry {
+     const char *key;
+     int value;
+   };
+   std::vector<LookupEntry> small_lookup = {
+     {"key1", 10},
+     {"key2", 20},
+     {"key3", 30},
+   };
+   // Or std::array if size is compile-time constant:
+   // std::array<LookupEntry, 3> small_lookup = {{ ... }};
+   ```
+   Linear search on small datasets is often faster than hashing or tree overhead, but this depends on
+   lookup frequency; in hot paths the O(1) vs O(n) difference may still matter. It is the heavy containers
+   (`map`, `set`, `unordered_map`) that should be avoided for small datasets unless profiling shows
+   otherwise.
+
+5. **Strings set once from configuration:** use `StringRef` (`esphome/core/string_ref.h`) rather than
+   `std::string`. Code generation passes a string literal that lives in flash for the life of the program,
+   so storing a `std::string` copies it onto the heap for nothing. `StringRef` is a non-owning pointer plus
+   length; it does not copy, and it must only ever refer to storage that outlives it (a string literal, or
+   a buffer owned elsewhere).
+   ```cpp
+   // Bad - heap copy of a literal that is already in flash
+   void set_keys(std::string keys) { this->keys_ = std::move(keys); }
+   std::string keys_;
+   ```
+   ```cpp
+   // Good - no allocation
+   void set_keys(const char *keys) { this->keys_ = StringRef(keys); }
+   StringRef keys_;
+   ```
+
+6. **Avoid `std::deque`:** it allocates in 512-byte blocks regardless of element size, guaranteeing at
+   least 512 bytes of RAM usage immediately. This is a major source of crashes on memory-constrained
+   devices.
+
+7. **Never use `new (std::nothrow)`:** on ESP-IDF exceptions are disabled, so a failed nothrow allocation
+   aborts instead of returning `nullptr`. Use `RAMAllocator` from `esphome/core/helpers.h`; CI rejects
+   `std::nothrow`.
+
+8. **Detection:** look for these patterns in compiler output:
+   - Large code sections with STL symbols (vector, map, set)
+   - `alloc`, `realloc`, `dealloc` in symbol names
+   - `_M_realloc_insert`, `_M_default_append` (vector reallocation)
+   - Red-black tree code (`rb_tree`, `_Rb_tree`)
+   - Hash table infrastructure (`unordered_map`, `hash`)
+
+Callback storage (`CallbackManager` vs `LazyCallbackManager`) and templated registration methods are
+covered in the `automations` skill.
+
+## Where to spend effort
+
+Prioritize container optimization for core components (API, network, logger), widely used components
+(mdns, wifi, ble), and components causing flash size complaints. Avoiding heap allocation after `setup()`
+is always required regardless of component type; the priority list is about effort spent on container
+migration (e.g. `std::vector` to `StaticVector`).

.agents/skills/hardware-test/SKILL.md

@@ -67,9 +67,17 @@ driven, what the logs show, and the same flow on a `dev` build when a quirk coul
    automations in the config itself.
 7. **Analyze** with `grep -a -c` counts over the capture and keep an `-events.txt` extract of the lines
    that matter. Decode crashes with the platform's `addr2line` (see `reference/boards.md`).
-8. **Report** on the PR as a table: build SHA, board and framework, what was driven, counts, result,
-   plus the `dev` control row and the exact flow in a `<details>` block. No credentials, no port names,
-   no IP addresses.
+8. **Report** on the PR. Every result comment starts with this two-line GitHub alert, then a blank line:
+
+   ```
+   > [!NOTE]
+   > Automated testing results
+   ```
+
+   Then one short sentence saying what was tested (builds and SHAs, board, what was driven), a results table
+   (build SHA, board and framework, what was driven, counts, result, plus the `dev` control row), short bullets
+   for the findings, and the setup and exact flow in a collapsed `<details>` block. Answer any question asked on
+   the PR directly in the comment. No credentials, SSIDs, IP addresses, port names or other bench details.
 
 ## Flashing by chip family
 
@@ -80,6 +88,7 @@ driven, what the logs show, and the same flow on a `dev` build when a quirk coul
 | ESP32-S3, C3, C6, H2 native USB | `esphome upload --device /dev/cu.usbmodem*` | `logger: hardware_uart: USB_SERIAL_JTAG`. The USB console can wedge when the app takes over the console (0 bytes, esptool `No serial data received`); replug, or watch through `esphome logs --device <name>.local` |
 | Pico W / Pico 2 W | `esphome upload --device <port>` (1200 baud touch); first flash ever via BOOTSEL and `--device /Volumes/RPI-RP2` or `RP2350` | port re-enumerates after upload: `ports.py --wait 2e8a` then capture |
 | BW15 kit (RTL8720CF, rtl87xx) | `~/.platformio/penv/.libretiny/bin/python $SKILL/scripts/flash_rtl87xx_kit.py flash write -d <port> <build>/.pioenvs/<name>/*.uf2`. Plain `esphome upload` and plain ltchiptool time out because opening the port asserts DTR/RTS and holds the chip in reset | `serial_capture.py --released`, never `esphome logs` |
+| BW12 and other AmebaZ boards (RTL8710BX, rtl87xx) on a plain USB-UART | `~/.platformio/penv/.libretiny/bin/python -m ltchiptool flash write -B 115200 -b 115200 -t 120 -d <port> <uf2>`, started first, then strap by hand: CEN to GND, TX2 to GND, release CEN, release TX2. Without `-B 115200` the system data read at 1.5 Mbaud fails on CH340s (`Data read error at 0x9000`) and nothing is written. A 1 MB image takes about 80 s | `serial_capture.py`; a reset pulse does not reach CEN on a plain adapter, so a capture starts mid-run |
 | BK7238 / CB3S (bk72xx, RTS on CEN) | `~/.platformio/penv/.libretiny/bin/python $SKILL/scripts/flash_bk72xx.py flash write -d <port> <uf2>` (about 75 s) | logs are on UART2 by default; `logger: hardware_uart: UART1` puts them on the flashing adapter. `serial_capture.py --reset` gives a chip-enable reset; a plain open may or may not reset it, check for the bootloader banner |
 | LN882H | ltchiptool forces 1 Mbaud for writes; weak adapters need `LN882H_YM_BAUDRATE` patched to 115200 in the penv's `ln882htool.py`; short BOOT to GND while the tool connects | OTA after the first flash |
 | nRF52840 Dongle (PCA10059) | Nordic `nrfutil nrf5sdk-tools dfu usb-serial` on the bootloader port, or `dfu: true` in the config so the 1200 baud touch pin-resets into the bootloader; adafruit-nrfutil is refused | app has two CDC ports, ESPHome log on the second |
@@ -114,8 +123,8 @@ Per-topic scripts, each described in the matching `reference/recipes/<component>
 (server with prefixed misbehaviour cases, PNG, JPEG, BMP and QOI generators, API drivers, first pass timer), `zwave_proxy/` (checksum verifier, device info), `usb_cdc_acm/`
 (host driver, Zephyr driver probe), `rtsp_server/` (protocol client, raw rate reader, measurement set,
 API image grab), `cse7766/` (API observer with ping), `zephyr_coredump/` (reader flow, decode with the
-Zephyr tools), `libretiny_prefs/` (rtc preference harness header), `libretiny_crash/` (BK72xx crash probe header and
-capture decoder), `api_outgoing/` (wake cycle and Wi-Fi drop timing readers), `ld2450/` (zone number driver, loopback harness driver with number and binary sensor states and TX frame summarizer), `api/` (raw plaintext native API client for user defined actions, store_yaml recovery client over Noise or plaintext), `http_request/` (update manifest server, update state driver, request test server, action case driver, concurrency, reboot, update check and OTA drivers, `ota_auth/` OTA request cases with basic auth), `modem/` (button and sensor driver over the plaintext API, also used for the BLE MIDI boards), `dfrobot_sen0395/` (API driver and case runner), `homeassistant_light/` (direct bulb driver, held-report case driver), `usb_audio/` (URL player), `speaker_media_player/` (API player driver, pipeline task case runner), `web_server/` (SSE client, request mix, SSE and HTTPD case runners), `mqtt/` (broker config with TLS, client-certificate and WebSocket listeners, test CA with server and client certificates, transport case runner, flash-and-drive runner, CLI TLS check without a board, MAC suffix runner with per-variant configs for two boards, `shutdown/` shutdown message runner and matcher), `udp/` (datagram recorder, `udp.write` payload checker and case driver), `entity_icons/` (icon and device class readback over the API, MQTT discovery, web_server and dump_config, with a comparer), `uart_sim/` (API driver and flash, capture and drive runner for the simulated UART sensors), `remote_base/` (loopback runner and API log decoder per sent frame).
+Zephyr tools), `libretiny_prefs/` (rtc preference harness header), `libretiny_crash/` (BK72xx crash probe header,
+capture decoder, RTL8720C ROM dump versus report matcher), `api_outgoing/` (wake cycle and Wi-Fi drop timing readers), `ld2450/` (zone number driver, loopback harness driver with number and binary sensor states and TX frame summarizer, frame stream harness with presence hold and link loss drivers, BT MAC driver), `api/` (raw plaintext native API client for user defined actions, store_yaml recovery client over Noise or plaintext), `http_request/` (update manifest server, update state driver, request test server, action case driver, concurrency, reboot, update check and OTA drivers, `ota_auth/` OTA request cases with basic auth), `modem/` (button and sensor driver over the plaintext API, also used for the BLE MIDI boards), `dfrobot_sen0395/` (API driver and case runner), `homeassistant_light/` (direct bulb driver, held-report case driver), `usb_audio/` (URL player), `speaker_media_player/` (API player driver, pipeline task case runner), `web_server/` (SSE client, request mix, SSE and HTTPD case runners), `mqtt/` (broker config with TLS, client-certificate and WebSocket listeners, test CA with server and client certificates, transport case runner, flash-and-drive runner, CLI TLS check without a board, MAC suffix runner with per-variant configs for two boards, `shutdown/` shutdown message runner and matcher), `udp/` (datagram recorder, `udp.write` payload checker and case driver), `entity_icons/` (icon and device class readback over the API, MQTT discovery, web_server and dump_config, with a comparer), `uart_sim/` (API driver and flash, capture and drive runner for the simulated UART sensors), `remote_base/` (loopback runner and API log decoder per sent frame), `xiaomi_ble/` (frame and header generator with the expected counts, injecting, advertising and receiving configs, counter and runner), `ble_terminate/` (proxy and peer configs, connect and disconnect driver, peer disconnect matcher, runner), `i2c_slave/` (D1 mini slave and BK7238 master configs), `wpa3_ap/` (PMF AP with join and leave logging, BK7238 and CB3S DUTs, runner and summary), `wisp_ap/` (WISP upstream with AP buttons, WISP and ALWAYS_ON DUTs, client, runner and per phase summary), `wifi_restart/` (WPA3 mixed WISP upstream, static IP DUTs with syslog, runner and summary), `ota_timing/` (OTA upload timer with deflate, raw and TCP modes, API build check, A/B runner and median summary, `inflate_bench/` on-device inflate benchmark with stream generator, configs and syslog runner), `gpio_hold/` (ISR pin hold self test with ESP32 and S3 configs), `ble_event_inject/` (GAP and GATTC event injector header and proxy package), `light_sleep_wake/` (LEDC wave and edge timer for light sleep wakeup on a C3), `manual_ip/` (station and AP manual IP configs, DNS probe and runners), `web_handlers/` (late handler, web_server, captive_portal and prometheus variants, web OTA runner), `bk72xx_flash/` (flash readback with retries, per partition diff of two dumps), `ble_server_security/` (secure GATT server and pairing client configs, driver, summary and runner), `prometheus/` (datetime metrics configs and driver).
 
 ## Templates in `$SKILL/templates`
 

.agents/skills/hardware-test/reference/boards.md

@@ -125,9 +125,26 @@ Ports are not listed on purpose: they follow the USB hub port and change on ever
 - The BK7238 bootloader prints `App verify failed! Need to recovery factory firmware.` on every boot with any
   build; it is harmless. Even with the right layout an ESP32 client did not see the BK7238 fallback AP on
   LibreTiny 1.13.0 or 1.14.0, so the wifisave hammer cannot run on it.
+- Moving a BK7231N device (CB3S) that sits on its fallback AP to another network: join the host's Wi-Fi to
+  `<name>-fallback`, OTA to 192.168.4.1, then rejoin the host's own network. Moving it with `/wifisave` instead can
+  leave it in the captive portal's BLIND_RETRY state (no scan, `Beacon Timeout` every 46 s) until it reboots, and a
+  log-only serial adapter cannot reset it. Rejoining may change the host's address; update a syslog target that
+  points at the old one.
 - Callbacks run on the `async_tcp` task and log lines can lose a byte. Wi-Fi connect takes tens of ms;
   radio enable 160 to 278 ms, disable about 200 ms.
 
+## BW12 and other AmebaZ boards (RTL8710BX, LibreTiny rtl87xx)
+
+- Download mode is a hand strap: CEN to GND, TX2 to GND, release CEN, release TX2. A plain USB-UART
+  does not drive CEN, so neither ltchiptool nor `serial_capture.py --reset` can reset the chip.
+- ltchiptool links at 1.5 Mbaud and reads the system data at 0x9000 at that rate before it writes at
+  115200. A CH340 drops bytes there (`Data read error at 0x9000`, `Timed out waiting for Quiet-Handshake`)
+  and the run gives up without writing. `-B 115200 -b 115200` moves the link to 115200 first and works.
+  Start the command with `-t 120` so it is already polling while the board is strapped.
+- After a failed run the ROM may sit at another baud rate and no longer link; strap again.
+- The device builder frontend's AmebaZ flasher follows ltchiptool (link and system read at 1.5 Mbaud),
+  so it can hit the same read failure on the same adapters.
+
 ## BW16 kit (RTL8720DN, AmebaD)
 
 - The kit's own USB-C (CH340) is on LOG_UART and flashes fine. Its two-transistor circuit resets while
@@ -172,6 +189,12 @@ Ports are not listed on purpose: they follow the USB hub port and change on ever
   the port open, and the logger warns `logger took a long time`. Print one line per `loop()` instead.
 - Zephyr settings survive a reflash: key a `restore_value` global by a `run` substitution when a test needs
   a never-saved value.
+- Thread: nRF52 `openthread:` accepts only `tlv:`; the individual keys (network_key, channel, ...) are rejected at
+  validation. To list nearby networks with channel and PAN ID, call
+  `otThreadDiscover(InstanceLock::acquire().get_instance(), 0, 0xFFFF, false, false, cb, nullptr)` from an
+  `interval:` lambda in an included header; results arrive on the OpenThread thread and log fine.
+- A wrong network key on a real network's channel and PAN ID makes the device form its own partition as leader on
+  that PAN within about 30 s; take it off Thread at once. Never try guessed datasets against a network in use.
 
 ## Waveshare ESP32-S3-ETH with the OV5640 camera module
 

.agents/skills/hardware-test/reference/recipes/README.md

@@ -11,20 +11,20 @@ One file per component, named after it:
 - [`espnow.md`](espnow.md): ESP8266 broadcast receive
 - [`hdmi_cec.md`](hdmi_cec.md): two-board wire
 - [`http_request.md`](http_request.md): update manifest URL changed at run time, every setter shape, a swap under a running check, non-blocking request actions with loop gap, queue, cancel and reboot checks, OTA request cases with basic auth and MD5 sources
-- [`ld2450.md`](ld2450.md): polygon zones, zone number round trip through the radar's readback, ESP8266 loopback harness without a radar
+- [`ld2450.md`](ld2450.md): polygon zones, zone number round trip through the radar's readback, ESP8266 loopback harness without a radar, presence hold and link loss on a frame stream, BT MAC readback
 - [`ld2412.md`](ld2412.md): output pin level through the basic configuration command, held back writes
-- [`libretiny.md`](libretiny.md): rtc preference emulation, BK72xx crash report on the next boot
+- [`libretiny.md`](libretiny.md): rtc preference emulation, BK72xx and RTL8720C crash report on the next boot with a ROM dump matcher, BK7238 BLE bond store safety with a full flash backup and per partition diff
 - [`modem.md`](modem.md): PPPoS modem component boot check with no modem attached, loop blocking and disable timing
 - [`mqtt.md`](mqtt.md): broker bench, QoS, reconnect, crash log over MQTT, WebSocket transports with a local and a public broker, TLS trust sources (own CA, client certificate, certificate bundle, no verification) with the CLI check, MAC suffix topics and will messages on two boards, shutdown message delivery on restart, disable and deep sleep
 - [`nrf52.md`](nrf52.md): dongle board support
 - [`online_image.md`](online_image.md): stalled and malformed downloads, PNG, JPEG, BMP and QOI decode time per loop pass with large images and small buffers
 - [`openthread.md`](openthread.md): border router on the S3 host plus H2 RCP
-- [`ota.md`](ota.md): overlap and lock, timeouts, deflate
+- [`ota.md`](ota.md): upload without a configuration and the refusal checks, overlap and lock, timeouts, deflate, upload timing A/B per build and board with deflate and raw, on-device inflate benchmark
 - [`rtsp_server.md`](rtsp_server.md): camera streaming on the Waveshare ESP32-S3-ETH
 - [`scheduler.md`](scheduler.md): loop wake A/B
 - [`udp.md`](udp.md): datagrams, multicast, sockets, `udp.write` payload shapes
 - [`usb_cdc_acm.md`](usb_cdc_acm.md): nRF52840 dongle host driver
-- [`wifi.md`](wifi.md): reboot_timeout with a fallback AP, DHCP lease cache, Easy Connect (DPP), WISP bridge with an ethernet upstream
+- [`wifi.md`](wifi.md): reboot_timeout with a fallback AP, DHCP lease cache, Easy Connect (DPP), WISP bridge with an ethernet upstream, reboots against a WPA2/WPA3 PMF access point, WISP and ALWAYS_ON AP switching, AP restarts with WPA3 mixed and LIGHT power save
 - [`zephyr.md`](zephyr.md): preferences on the nRF52840 dongle
 - [`zephyr_coredump.md`](zephyr_coredump.md): crash and decode on nRF52 boards
 - [`zwave_proxy.md`](zwave_proxy.md): frame checksums on a live proxy
@@ -36,5 +36,14 @@ One file per component, named after it:
 - [`web_server.md`](web_server.md): state SSE stress and the HTTPD task stack on web_server_idf, entity icon and device class readback
 - [`uart_sim.md`](uart_sim.md): UART sensors simulated on a classic ESP32 with loopback and cross-wired UARTs
 - [`gpio_expander.md`](gpio_expander.md): interrupt driven expander inputs, loop idle checks, testing a device out of reach
-- [`led_strip.md`](led_strip.md): forcing a failed strip allocation, RMT symbols on the S3
+- [`led_strip.md`](led_strip.md): forcing a failed strip allocation, RMT symbols per chip, loop cost per strip
 - [`remote_base.md`](remote_base.md): remote_receiver protocol decoding through a one-pin loopback on ESP32 and ESP8266, no IR hardware
+- [`xiaomi_ble.md`](xiaomi_ble.md): Xiaomi, ATC and PVVX parsing by injecting scan results into the ESP32 tracker or advertising them from a second ESP32, real AES-CCM frames, duplicate filter model
+- [`ble_terminate.md`](ble_terminate.md): proxy disconnect against a peer with peripheral latency
+- [`i2c_slave.md`](i2c_slave.md): fake I2C slave on a D1 mini, ESP32 hardware master as the control
+- [`gpio_hold.md`](gpio_hold.md): GPIO hold through the ISR pin API, pad read back with no wiring, per chip pin notes
+- [`ble_event_inject.md`](ble_event_inject.md): made up GAP and GATTC events fed into the ESP32 BLE handlers from a button
+- [`light_sleep_gpio_wake.md`](light_sleep_gpio_wake.md): GPIO wakeup from auto light sleep with an LEDC wave that keeps running in sleep
+- [`manual_ip_and_web_handlers.md`](manual_ip_and_web_handlers.md): station, AP and Ethernet manual IP, web handlers added late and web OTA with web_server, captive_portal or prometheus
+- [`ble_server_security.md`](ble_server_security.md): esp32_ble_server pairing, encrypted and MITM characteristics, subscription before and after pairing, bond reuse per ESP-IDF version
+- [`prometheus.md`](prometheus.md): date, time and datetime metrics in UTC and another time zone

.agents/skills/hardware-test/reference/recipes/ble_event_inject.md

@@ -0,0 +1,20 @@
+# Injecting BLE GAP and GATTC events (#21698)
+
+Some BLE paths (repeated connection parameter update failures, a disconnect in the middle of them) are hard to get
+from a real peer on demand. `scripts/ble_event_inject` feeds made up events straight into the ESP32 BLE handlers.
+
+- `ESP32BLE::gap_event_handler` and `gattc_event_handler` are protected statics and the class is `final`.
+  `cp_inject.hpp` reaches them through explicit template instantiation, where access checks do not apply. Its
+  `cpinject::run()` sends six labelled steps (`STEP1` to `STEP6`): four peers failing five times each, a success
+  followed by more failures, and disconnects between failures.
+- `cp-inject.yaml` is a package for a copy of a proxy config: it adds the include, a VERBOSE logger with the busy
+  tags back at DEBUG, and an `Inject conn param events` button. Press it with `scripts/api_button.py` while a
+  log capture (`esphome logs ... --device <ip>` under `timeout`) runs, then count the warning and verbose lines per
+  step on the PR and on dev.
+- The handlers push into a queue and event pool that take a single writer, normally the Bluetooth task. Injecting
+  from the main task while real events arrive can corrupt them, so the button stops the scan, waits for it to end,
+  injects and starts it again. Run it with no active proxy connections (no Home Assistant BLE client holding one).
+- Fake disconnects use `gattc_if` 0xFE: every client and proxy connection filters on its own `gattc_if_`, so the
+  queued event reaches only the shared handler. Do not use 0xFF (`ESP_GATT_IF_NONE`), which every client accepts.
+- The fake peers (`C0:FF:EE:00:00:0n`) are never connected, so a live proxy keeps serving its real connections;
+  put it back on its own config afterwards.

.agents/skills/hardware-test/reference/recipes/ble_server_security.md

@@ -0,0 +1,26 @@
+# esp32_ble_server pairing and encrypted characteristics (#17562)
+
+`scripts/ble_server_security`: a classic ESP32 devkit runs the secure GATT server (`server.yaml`) and a second ESP32
+(an S3 here) runs `ble_client` as the central (`client.yaml`). The server has a `read_encrypt` battery level
+(`99.9`), a plain characteristic, a `write_encrypt` characteristic that logs what it receives, a `read_encrypt_mitm`
+characteristic and a `read_encrypt` characteristic that notifies every 2 s. `run.sh TAG IO AUTH` builds both with the
+pairing mode as substitutions, pairs once, reconnects twice, and `summary.py` prints PASS or FAIL. The server needs
+`esp32_ble_server` security from #17562 or later: point `SERVER_TREE` at that checkout (the client builds from
+`TREE`, dev is fine). `summary.py` needs each refused read to show an `Error reading char` reply, so a run where the
+client never connected fails instead of passing.
+
+- `ble_client` never starts pairing on its own (a phone does, after an `insufficient authentication` reply); the
+  client's `Pair` button calls `BLEClientBase::pair()`. Read errors show as `Error reading char at handle N, status=5`
+  (not authenticated: no bond) or `status=15` (not encrypted: bond known, link not encrypted yet).
+- The `ble_client` text sensor subscribes once, at connect, before any pairing, so a protected characteristic refuses
+  it. The `Subscribe` button writes the CCCD again after pairing; `ble_client` drops its service list after discovery,
+  so it writes the characteristic handle + 1 (where ESP-IDF's GATT server puts the CCCD).
+- Numeric comparison (`display_yes_no` with `sc_mitm_bond`) tells a fresh pairing from a bond reuse: the
+  `numeric comparison` line only appears on a fresh pairing. Just Works cannot show it.
+- ESP-IDF 5.5.6 drops the bond on the second disconnect on both roles (`Deleting unpaired device`), so that reconnect
+  pairs from scratch; 5.5.5 keeps it (`-s idf 5.5.5`).
+- The server keeps its bonds in NVS. One holding an authenticated bond for the client refuses a Just Works re-pairing
+  silently; erase the devkit's flash before switching from nc to jw.
+- `summary.py` failed both modes at 58636914ab: `read_encrypt_mitm` was readable with no pairing, because
+  `BLEService::create_characteristic()` took an 8 bit `esp_gatt_char_prop_t` and dropped the MITM bits (1 << 8 and
+  1 << 9). Check the generated `create_characteristic` call and the property type when a flag above bit 7 is added.

.agents/skills/hardware-test/reference/recipes/ble_terminate.md

@@ -0,0 +1,27 @@
+# BLE terminate with peripheral latency (#20232)
+
+`scripts/ble_terminate`: an ESP32-S3 `bluetooth_proxy` (active) over Ethernet as the central (`btproxy.yaml`) and a
+classic devkit `esp32_ble_server` peer (`btpeer.yaml`). The peer asks for 50 ms / latency 10 / 6 s with
+`esp_ble_gap_update_conn_params` 4 s after connect; earlier requests are overwritten by the proxy's own "medium
+conn params" after service discovery. Read the proxy's `Conn params updated` line (VERBOSE, API logs) to confirm;
+the peer side never logs the event. The proxy only connects to a device its scanner has seen, so `btdrive.py`
+subscribes to raw advertisements first.
+
+Measure the peer's own disconnect (`serial_capture.py --ms`) after the client's disconnect: about 0.2 to 0.8 s when
+the terminate arrived, about 6 s (the supervision timeout) when it was lost. The client side disconnect time is the
+API reply and says nothing. On dev the S3 lost 6 of 12; the PR lost 0 of 16.
+
+`run.sh LABEL ROUNDS` writes `central_bda.hpp` from `PROXY_BT_MAC`, builds and flashes both boards, runs the rounds
+and `an.py`, which prints PASS only when every round completed, every peer disconnect came within 3 s and the last
+parameter update the proxy logged before each disconnect had latency=10 (a round without it did not test peripheral
+latency and counts as inconclusive).
+
+Controller build check: ESP-IDF 5.5.6 ships `libbtdm_app.a` byte for byte the same as 5.5.5 for esp32c3 and esp32s3
+(build 51d9dfd). Without hardware:
+
+```bash
+strings -a <idf>/components/bt/controller/lib_esp32c3_family/<chip>/libbtdm_app.a | grep -E '^[0-9a-f]{7}$'
+```
+
+On a board: a button lambda logging `btdm_controller_get_compile_version()` (declared `extern "C"`) over the API,
+since the ESP-IDF init line only reaches the UART.

.agents/skills/hardware-test/reference/recipes/gpio_hold.md

@@ -0,0 +1,17 @@
+# ESP32 GPIO hold through the ISR pin API (#21576)
+
+`scripts/gpio_hold/hold_test.hpp` builds `ESP32InternalGPIOPin`s by hand, calls `to_isr()` and reads each pad back
+with `gpio_get_level`: 20 ISR writes, a raw `gpio_set_level` control (a held pad must not change, an unheld one must)
+and ISR `pin_mode` pulls (input must drop to `NONE` between pull changes, because the ISR path only applies pulls when
+`FLAG_INPUT` toggles). It needs no wiring: unconnected pins read the pulls. Each pin runs held and unheld, then the
+round logs `ROUND n PASS failures=0` or `FAIL`.
+
+- `USE_GPIO_HOLD` comes from any pin with `hold_state: true`, so the configs add a `gpio` output on a spare pin.
+- `gpio-hold-esp32.yaml` (classic devkit) runs three rounds 5 s after boot and has a `rerun` button;
+  `gpio-hold-esp32s3.yaml` (Waveshare ESP32-S3-ETH, where most pins go to the W5500, TF card, camera, WS2812 or
+  USB) runs a round every 15 s.
+- C3: ESP-IDF 5.5 has `SOC_RTCIO_PIN_COUNT` 0, so `rtc_gpio_is_valid_gpio()` is false on every pin and only the
+  digital hold path runs.
+- C6 (Waveshare ESP32-C6-GEEK, native USB): LP pads GPIO0 to 7 are RTC IO with hold. GPIO1, 2, 7, 20, 23 read high
+  and GPIO6 low whatever the pull (LCD, TF card or resistors); clean pins are 0, 3, 4, 5, 18, 19, 21, 22. Dev
+  already holds LP pads through `gpio_hal_hold_en`, so the PR and dev look the same there. H2 is untested.

.agents/skills/hardware-test/reference/recipes/i2c_slave.md

@@ -0,0 +1,12 @@
+# Fake I2C slave (#21586)
+
+`scripts/i2c_slave`: a D1 mini running an Arduino `Wire` slave at 0x40 (`i2cslave.yaml` with `i2cslave.hpp`): a fake
+HDC1080 plus a 256 byte register file, and reads of 0x7E and 0x7F stretch the clock for 0.5 and 5 ms. It needs
+`libraries: [Wire]` and 160 MHz. With jumper wires only and the internal pull ups it works to about 50 kHz; it cannot
+keep up with a true 100 kHz hardware master.
+
+The master (`i2cbk.yaml`, a BK7238 on P26/P24, needs the software I2C bus from that PR) writes a counter, reads it
+back with a repeated start, probes an absent address and reads both stretch registers every 2 s, logging `RW OK`
+(the read back matches, the absent address NACKs and both stretched reads succeed) or `RW BAD` with running totals.
+Control: the same blocks on a classic ESP32 devkit (SDA GPIO21, SCL GPIO22) with hardware I2C at 25 and 50 kHz got
+every repeated start read right, so failures on a bit banged master at a similar real clock point at the master.

.agents/skills/hardware-test/reference/recipes/ld2450.md

@@ -40,3 +40,29 @@ least 24 selects per build and read the raw bytes with `uart: debug: direction:
 id) before blaming the parser; the API log stream drops lines under that dump volume, so some misses stay
 unclassified.
 Results: https://github.com/esphome/esphome/pull/21357#issuecomment-6333244886
+
+## Frame stream harness: presence hold and link loss (#21733, #21734)
+
+`scripts/ld2450/stream-harness.yaml` is a loopback harness that feeds the hub one periodic frame every 100 ms from a
+mode global set by API buttons: no target, target 1 approaching, silent, or 12 garbage bytes per tick (what a radar
+at another baud rate looks like). `has_target` has `filters: []` so the hub's own hold is what gets timed, and every
+change logs `presence raw ON|OFF at <millis>`. Add `stream-connected.yaml` as a second package on builds whose ld2450
+has the link watchdog (`connected` binary sensor).
+
+- `holddrive.py HOST OUT`: for `presence_timeout` 5 and 0, three seconds of target frames, then empty frames; prints
+  the ms from the press to the off line, then sets 0 and 20 and reads the number back after a restart. On dev 0
+  still held 5 s and snapped back to 5; with #21733 it clears on the next empty frame (under 200 ms) and stays 0.
+- `linkdrive.py HOST OUT`: steps through target frames, 10 s of silence, empty frames, 10 s of garbage, target
+  frames, then the hub's radar restart followed by 3.5 s and by 15 s of silence; prints every state change and link
+  log line with its time. Dev freezes every entity through silence and garbage. Run against a real radar too: a
+  radar restart and a Bluetooth toggle must not report a loss (the Apollo R PRO-1 ETH did not).
+- A freshly booted ESP8266 delays API commands sent in its first seconds and then runs them back to back; the
+  drivers wait 8 s after connecting before timed presses.
+
+## BT MAC sensors (#21731)
+
+`scripts/ld2450/macdrive.py HOST OUT` reads every LD24xx BT MAC, Bluetooth and firmware entity, turns each Bluetooth
+switch off, then on, and prints the values after each step. The MAC sensors are often `disabled_by_default` in
+vendor configs but still publish over the API. On the Apollo R PRO-1 ETH both radars (LD2450 fw 2.04, LD2412 fw
+1.26) have the multicast bit set in the first MAC byte, which dev showed as `unknown`. A radar MAC identifies a
+device: quote only its first byte in PR text.

.agents/skills/hardware-test/reference/recipes/led_strip.md

@@ -8,3 +8,7 @@
 - On the ESP32-S3 give each RMT strip `rmt_symbols: 48`, or the first channel takes all 192 symbols.
 - A board with a few WS2812 on a pin already defined as a light can be extended with `id: !extend <light id>` to
   add effects; restore its own config afterwards.
+- Loop cost of a strip (#19337): give each strip an `addressable_lambda` effect with `update_interval: 0ms` that
+  changes every pixel, and add `runtime_stats: log_interval: 30s`; compare the main loop and light times per pass,
+  leaving out the first period after boot. The strips need no LEDs attached. A classic ESP32 has 512 RMT symbols in
+  all, so three strips need `rmt_symbols: 128` each.

.agents/skills/hardware-test/reference/recipes/libretiny.md

@@ -56,3 +56,46 @@ OTA-only board: copy the device's own config, add the probe include and the step
 next to that config. A LibreTiny OTA is a 1.5 MB UF2 and takes minutes on a slow link, so do not wrap the
 upload in a short timeout. Read the replayed report with `esphome logs ... --device <name>.local` (an API
 client gets it on subscribe) or on the board's serial log port if it has one.
+
+## RTL8720C crash report on the next boot (#21648)
+
+`templates/rtl8720c-crash-buttons-dut.yaml` has one API button per case: `trap` (`__builtin_trap()`, usage fault),
+`bus fault` (store to 0x70000000, imprecise, CFSR 0x400), `wild call` (call to 0xFFFFFFF1, MemManage fault with PC
+0xFFFFFFF0 and LR in the caller), `isr trap` (a trap from a spare interrupt pended by software, hard fault), `hang`
+(interrupts off, the watchdog resets it and nothing must be reported) and `restart`. Every boot logs
+`boot reason N, crash record yes|no|n/a`; `n/a` means the handler is not built, as on a dev control build.
+
+Flow per build: flash with `flash_rtl87xx_kit.py`, start one `serial_capture.py PORT 600 OUT --released --ms`, then
+`scripts/libretiny_crash/rtl_drive.py HOST OUTDIR TREE YAML trap restart "bus fault" "wild call" "isr trap" hang restart`.
+The `restart` after the first crash checks the report shows only once. Time each reset from the serial capture
+(the ROM fault dump to the next `ROM Version` line), not from the API port: the board is back before an API client
+notices it went away. Compare the reported PC with the `PC  =` line of the ROM dump, and check the decoded lines in the
+`caseN-*.txt` files. Pass the UF2 by name (`firmware.uf2`): the build directory holds a second identically sized copy.
+
+Regression rerun of the same flow (#21728): `scripts/libretiny_crash/rtl_rom_vs_report.py CAPTURE` lists, for each ROM
+fault dump in a `--ms` capture, the ROM's CFSR and PC, the ms to the next `ROM Version` banner and the Reason, CFSR, PC
+and SP the next boot reported, so two builds compare line by line. The ROM lines start with a bare CR; split the
+capture on newlines only. The imprecise bus fault's PC moves between builds (it is wherever the core was when the
+fault was raised); compare it with that run's ROM dump, not with another build. SP stays the same per fault (8 word
+frame, no FPU). On the BK7238 time the reboot from the button press to the next `BK7238_1.0.14` banner (0.6 to 0.7 s).
+
+## BK7238 BLE and flash sector safety (#21735)
+
+The Beken BLE bond store reads its partition at boot and erases that sector when the CRC does not match; before
+LibreTiny 1.14 the partition was unmapped and pointed at sector 0, erasing the bootloader (#18646). To prove a change
+like this safe on a board that cannot be recovered without a dump:
+
+1. Take a full backup first: `scripts/bk72xx_flash/readback.sh full-before` (ltchiptool family `beken-7238`, about
+   190 s for 2 MB). Keep it until the board is back on a known build.
+2. Serial flash a build without the feature, then deliver the feature build by OTA (the path users take). Use
+   `bk72xx: framework: sdk_silent: none` with `logger: hardware_uart: UART1`, so the SDK's own lines (bootloader
+   banner, `CRC fail, remove bonded info`) show up in the capture.
+3. Soak with Wi-Fi up, OTA again from the new build, pulse CEN a few times (`serial_capture.py --reset`), and check
+   each boot comes back through the bootloader.
+4. Read the whole chip again and run `scripts/bk72xx_flash/partdiff.py full-before.bin full-after.bin
+   <libretiny>/boards/_base/beken-72xx.json`: bootloader, calibration and userdata must be unchanged; app, download
+   and one KVS sector change normally.
+
+Check what already sits in the target partition before the first boot: on an XH-WB3S that shipped with Tuya firmware,
+0x1E1000 (the `ble_bonding` partition of the generic Beken layout) held the Tuya `diff2ya` record (magic `ba dc cd
+ab`), which the bond store erased on its first boot. Nothing depended on it there, but say so in the results.

.agents/skills/hardware-test/reference/recipes/light_sleep_gpio_wake.md

@@ -0,0 +1,16 @@
+# GPIO wakeup from auto light sleep (#21074)
+
+`scripts/light_sleep_wake` needs no wiring. LEDC on `LEDC_USE_RC_FAST_CLK` with
+`sleep_mode = LEDC_SLEEP_MODE_KEEP_ALIVE` keeps toggling GPIO10 at 2 Hz through light sleep, and an
+`ESP32InternalGPIOPin` interrupt (`ANY_EDGE`) on the same pad timestamps every edge in an IRAM handler.
+`wakeprobe::report()` logs every 30 s: edges seen and expected, the shortest and longest gap, how many edges came
+more than 20 ms off the half period, and the SLEEP share from `esp_pm_dump_locks()` (needs `profiling: true`).
+
+- Native ESP-IDF builds need `include_builtin_idf_components: [esp_driver_ledc]`.
+- After `ledc_channel_config` call `gpio_input_enable(pin)`, never `gpio_set_direction(INPUT_OUTPUT)`: that routes the
+  pad back to a plain GPIO output, the wave stops and the pad reads a constant 1.
+- The C3 shares one LEDC clock across timers, so a board whose own config drives LEDs through `ledc` needs a test
+  config without them. Logging over the API works with `logger: baud_rate: 0`.
+- Read the reports over the API with `esphome logs ... --device <ip>` under `timeout`. Run PR, dev, then the PR
+  again. On a C3 bulb dev caught every edge but handled 8 of 845 more than 20 ms late
+  (up to 45 ms), the PR none; both slept 57 to 58% of the time.

.agents/skills/hardware-test/reference/recipes/manual_ip_and_web_handlers.md

@@ -0,0 +1,31 @@
+# Manual IP and web handler flows (#21708, #21709, #21711, #21712)
+
+## Station manual IP (`scripts/manual_ip/run-sta.sh`)
+
+Set `static_ip` to the board's own DHCP lease so nothing collides, and DNS2 to a server DHCP does not hand out
+(9.9.9.9), so the dump and `dns_getserver()` prove the manual values are in use. A missing SSID with its own
+`manual_ip` comes first, so the real network is a later table row and inherits the top level `manual_ip`. The probe
+logs both DNS servers and resolves a name every 20 s; `ipaddr_ntoa` uses one static buffer, so it uses
+`ipaddr_ntoa_r` to log two addresses in one line. PASS needs the address, both DNS servers, a lookup and a ping.
+
+## AP manual IP (`scripts/manual_ip/run-ap.sh`)
+
+The DUT has only a missing SSID, `ap_timeout: 10s` and `captive_portal:`, so its AP comes up on 10.42.7.1; the other
+board joins as a client and fetches `http://10.42.7.1/` every 10 s. A D1 mini AP drops some client requests on dev
+too, so compare failure counts with a dev run and do not expect zero.
+
+## Ethernet manual IP
+
+Same idea on an Ethernet board: its lease as `manual_ip` plus an outside DNS2 and the DNS probe from
+`scripts/manual_ip/dnsprobe.hpp`; restart it over the API and check again.
+
+## Web handlers (`scripts/web_handlers/run.sh`)
+
+`webbase.yaml` adds a GET /probe handler from `on_boot` at priority -100 (the shape of an external component adding
+a handler late) and takes a variant: `ws` (web_server), `cp` (captive_portal only, started on the station network by
+a button calling `global_captive_portal->start()`) or `prom` (captive_portal and prometheus). The runner fetches
+/probe and /, then uploads the same image through web OTA (`curl -F update=@<bin> http://ip/update`; ESP-IDF
+`build/<name>.bin`, not `firmware.ota.bin`; ESP8266 `.pioenvs/<name>/firmware.bin`) and checks over the API that the
+board is back on that build. On dev too, prometheus GET / gets no answer on ESP32 and 404 on ESP8266.
+
+A background `(cd dir && cmd) &` does not change the parent shell's directory: give capture scripts absolute paths.

.agents/skills/hardware-test/reference/recipes/nrf52.md

@@ -4,3 +4,10 @@
 
 Test config: the two LEDs as inverted GPIO outputs, the button as an inverted pull-up binary sensor, a
 10 s alive interval. To retest the nrfutil download path, move the cached `sdk-nrf/nrfutil` aside first.
+
+## ItsyBitsy over UF2 when serial DFU fails (#20481)
+
+Serial DFU to the ItsyBitsy nRF52840 fails about half the time ("Target is not in DFU mode"). Reliable path: a
+1200 baud touch on the app port, wait for the `ITSY840BOOT` volume, convert `zephyr/zephyr.hex` with the SDK's
+`zephyr/scripts/build/uf2conv.py -f 0xADA52840 -c` and copy the .uf2 to the volume (the copy ends in an I/O error
+when the board reboots, which is normal). macOS BLE scans give UUIDs, not MAC addresses.

.agents/skills/hardware-test/reference/recipes/ota.md

@@ -6,3 +6,64 @@ All overlap orders (a web upload during a native OTA and the reverse) must be re
 after an interrupted web upload waits about 14 s (ESP8266) or 7 s (ESP-IDF) for disconnect detection. A
 refused web client on `web_server_idf` never gets the busy reply; the connection drops at reboot.
 `pkill -f "a\|b"` does not alternate; use two pkills.
+
+## OTA upload timing A/B (#21639, #21641, #21642)
+
+`scripts/ota_timing/run.sh N PHASE...` times native OTA uploads per build (phase) and per board. It builds every
+board from `TREE_<PHASE>`, does one warm-up upload per board (it installs the phase's build and is left out of the
+results), checks over the API that each board reports the compile time from that build's `build_info.json`
+(`apiinfo.py`), then runs N rounds of every mode on every board. Boards are interleaved per round so they share the
+same air; put a nearby ESP32 on the same Wi-Fi in `BOARDS` as a network control. `summ.py` prints medians: never
+judge from single runs, slow Wi-Fi gave 33 to 235 s for the same upload in one night.
+
+- `otatime.py` drives `espota2.run_ota()` from the tree on `PYTHONPATH` and prints one `RESULT` line with the
+  `Update took` split, the compressed size, whether UDP was used and the time at each tenth of the upload.
+  An upload that leaves no `RESULT` line (board gone, script crash, timeout) is recorded as `rc: -1`, and `summ.py`
+  fails when a board and mode has fewer than N rows or an upload time cannot be read.
+- Boards with an `api: encryption: key` also encrypt OTA: set `API_KEY` for `otatime.py` and `apiinfo.py`. LibreTiny
+  LN882H listens for OTA on 8820, not 8892; BK72xx and RTL87xx use 8892.
+- Raw mode leaves `CLIENT_FEATURE_SUPPORTS_DEFLATE` out of the client's offer. Never switch deflate off after the
+  offer (`if deflate and 0:`): once the device accepted deflate it expects a deflate stream.
+- `-tcp` modes make `espota2._loss_reason` return None, so the client asks to stay on TCP. Clearing
+  `CLIENT_FEATURE_SUPPORTS_UDP` instead desyncs the handshake (`receiving UDP token response: timed out`).
+- LibreTiny trimmed UF2 (#21642): the CLI sends what `esphome.components.libretiny.uf2.device_ota_uf2()` returns for
+  `firmware.uf2`; that module only exists with #21642, so on older trees the full UF2 goes out. Compare the sizes
+  in the `Compressed to` lines, and rely on the API compile time check to show that the trimmed image booted.
+
+`receiving chunk result: Unknown error from ESP` is error 0xFF sent by the device, not a dropped connection. The
+device sends it for a decode or flash write error, and also when no data arrives for 105 s, so on a weak link it
+usually means a stall. A failure about 105 s after the last progress, or the same error on a raw upload, points to
+the link; to be sure, capture the device log (`Inflate err`, `Inflate overrun`, `No data received`).
+
+Results: https://github.com/esphome/esphome/issues/21639#issuecomment-6409628227 (dev and LibreTiny 1.14.0 on a
+CB3S), https://github.com/esphome/esphome/pull/21642#issuecomment-6412972562 (trimmed UF2 on CB3S, BK7238 and BW15),
+https://github.com/esphome/esphome/pull/21641#issuecomment-6412976859 (new decoder on the same three).
+
+## On-device inflate benchmark (#21639, #21641)
+
+`scripts/ota_timing/inflate_bench`: `gen_streams.py NAME=IMAGE...` cuts a 128 KB slice from each image and writes
+`bench_streams.hpp` with raw deflate streams (level 9, 4 KB and 32 KB windows). `bench.hpp` inflates every stream with
+each ring size at least the encoder window, three times, and logs the best time plus input and output rates, then
+`BENCHDONE`. `bench.sh CONFIG DEVICE` builds and uploads `bench-esp32s3.yaml` or `bench-bk72xx.yaml`, collects one pass
+over syslog and fails when a case did not reach the end of its stream.
+
+- LibreTiny defines `micros` as a macro, so the header calls `(esphome::micros)()`.
+- Keep the streams near 130 KB on a BK7231N: an image with about 330 KB of extra incompressible data failed its
+  OTA with "Writing OTA data to flash memory failed". Use one image and `--windows 12` there.
+- `bench.hpp` targets the decoder interface with `dest_limit`. #21641 drops it (the output runs into the window up
+  to `dest`); after that lands, remove the `dest_limit` line.
+
+Results: https://github.com/esphome/esphome/issues/21639#issuecomment-6409786224
+
+## Upload without a configuration (#21738)
+
+`esphome upload --file IMAGE --device IP` with no YAML picks the OTA port from the image header and takes the password
+from `--ota-password` or `ESPHOME_OTA_PASSWORD` and the key from `ESPHOME_OTA_KEY`. Test each chip family with its own
+current image (`build/firmware.ota.bin` for ESP-IDF, `.pioenvs/<name>/firmware.bin` for ESP8266, `firmware.uf2` for
+LibreTiny) and check the build over the API; switch one board between two builds to show a new image lands.
+
+Point every case that should be refused before connecting at an unroutable address (192.0.2.1), never at a board: a
+refusal prints no `Connecting to` line, and an image that slips through cannot reach a device. An ESP32-S3
+`firmware.factory.bin` slipped through on 2026-10-27 (its bootloader header at offset 0 looks like an app), the device
+reported `OTA successful` and its bootloader rejected the slot at boot; send the board its proper app image afterwards.
+A factory image has the partition table magic `AA 50` at 0x8000 and the app at 0x10000.

.agents/skills/hardware-test/reference/recipes/prometheus.md

@@ -0,0 +1,9 @@
+# prometheus
+
+## Date, time and datetime metrics (#9941)
+
+`scripts/prometheus/prom-datetime.yaml` is a D1 mini with `web_server`, `prometheus` and three optimistic template
+datetime entities; `prom-datetime-tz.yaml` adds an sntp time component with `HST10` (the zone applies without a sync).
+`datetime_metrics.py HOST` prints the rows of `/metrics`, sets the entities over the API to 2040 values (past 2038 on
+purpose) and prints them again with their UTC reading. Dates are local midnight in ms, times ms since midnight,
+datetimes ms since the epoch, so a datetime set as an instant reads the same in both zones.

.agents/skills/hardware-test/reference/recipes/wifi.md

@@ -37,3 +37,44 @@ AP start (the fix in that PR), the first AP start failed on every boot with `esp
 (DHCP server already started by the AP start event) and `Failed to enable NAT`, and the bridge only came up on the
 retry 90 s later; a client that cannot find the SSID for the first minute on an older build is that. With the fix
 the bridge is up 6 s after boot. Any change to the ESP-IDF AP start should also rerun the captive portal hammer.
+
+## Reboot against a WPA2/WPA3 PMF access point (#21590)
+
+- AP: `scripts/wpa3_ap/wdap.yaml` on a classic devkit. ESPHome `wifi: ap:` only, then an on_boot lambda switches it
+  to `WIFI_AUTH_WPA2_WPA3_PSK`, PMF capable, and registers a WIFI_EVENT handler that logs `STA JOIN` and
+  `STA LEAVE ... reason N` (the include header pulls in `esp_wifi.h`). It disables Wi-Fi after 420 s.
+- AP leave reason 3 means the station sent a deauth; reason 2 means the AP found a stale association when the
+  station came back; several reason 2 before one join are rejected reconnects.
+- DUTs (`wdbk.yaml` BK7238, `wdcb.yaml` CB3S, both on `wd-common.yaml`) list the AP at priority 10 and the house
+  network at 0, and reboot 30 s after joining the AP and 150 s after joining the house network. So they move to the
+  AP while it is up and fall back to the house network (where they can be flashed over the air) once its window
+  ends. `run.sh BUILD` flashes the AP, does OTA for the CB3S and a serial flash for the BK7238, then captures one AP
+  window; `an.py` prints each station's join and leave sequence and fails on a rejected reconnect, an AP that did
+  not switch to WPA2/WPA3 (`AP CONFIG FAILED`), a DUT missing from the AP events (`DUT_MACS`) or a BK7238 that never
+  joined the AP.
+- Gotchas: the LibreTiny build directory has two `.uf2` files, so flash `firmware.uf2`. The BK7238 serial flash
+  sometimes fails to link while the chip is in a reboot loop; retry it 3 or 4 times. OTA fails if the DUT reboots
+  mid-upload; retry.
+- Finding: BDK 3.0.78 rejects about 2 reconnects per reboot when the station left without a deauth; BDK 3.0.33
+  does not.
+
+## WISP and ALWAYS_ON AP rig (#20077)
+
+`scripts/wisp_ap`: the upstream is an ESP32-S3 with Ethernet running a WISP bridge (`wisp-upstream.yaml`) with its AP on
+10.40.0.1/24 and web buttons `ap off` and `ap on`. Its subnet must not be 192.168.4.0/24, the DUT AP's default, or
+nothing routes. The DUT is a classic devkit (`dut-wisp.yaml` or `dut-always-on.yaml`); the client is a D1 mini on
+the DUT's AP (`wispcli.yaml`) logging its DNS server, a name lookup, a GET by name, the AP itself and a LAN file by
+IP. `run.sh DUT START SECONDS T:STATE...` switches the upstream AP at the given times and `sum.py` groups the client
+results by phase. Capture the D1 mini without `--reset` (a CH340 floods garbage after an RTS reset). Clients keep
+the captive portal's DNS answer for its TTL.
+
+## AP restart with WPA3 mixed and LIGHT power save (#21531)
+
+`scripts/wifi_restart`: the WISP upstream (`wisp-upstream-wpa3.yaml`) switched to WPA2/WPA3 mixed by a script that
+runs again after every `ap on`, DUTs (`rebk.yaml`, `recb.yaml` on `re-common.yaml`) with a static IP on the bridge's
+subnet, a fallback AP, `power_save_mode: LIGHT` and syslog to the host through the NAT. `run.sh BUILD` switches the
+AP off and on three times; `an.py` fails on a reboot, a crash or a DUT that is not back on syslog within 120 s of an
+AP start. After 19 minutes the DUTs run `wifi.configure` to the house network; that keeps the static IP, so the next
+OTA needs the host to have an address on the bridge's subnet as well (an interface alias). The CB3S log is on UART1
+(TX1), so its configs need `logger: hardware_uart: UART1`. Runners must `pkill` only their own ports; a global pkill
+stops the other runs' captures.

.agents/skills/hardware-test/reference/recipes/xiaomi_ble.md

@@ -0,0 +1,38 @@
+# Xiaomi BLE
+
+## Xiaomi, ATC and PVVX parsing without the sensors (#21483)
+
+No Xiaomi devices are needed, and one ESP32 cannot hear its own advertising, so the frames are injected into the
+tracker as scan results: everything after the radio runs as normal.
+
+- `scripts/xiaomi_ble/gen.py` writes `frames.hpp` and `expected.json`: plain MiBeacon (0x50), encrypted v5 23 byte
+  with MAC and 19 byte without, real AES-CCM (`cryptography` AESCCM, tag 4, AAD 0x11, nonce = MAC reversed + pid +
+  cnt + 3 byte ext counter), ATC (13 byte, big endian) and PVVX (15 byte, little endian) 0x181A frames, a corrupted
+  tag every other cycle and a counter 0 boot frame. It models the old shared duplicate counter (`base`) and the per
+  device one (`pr`) and prints the expected publish counts.
+- `xbinject.yaml` (ESP-IDF devkit) includes `frames.hpp`; an `interval:` builds a `BLEScanResult` and calls
+  `global_esp32_ble_tracker->gap_scan_event_handler(r)`. `count.py` counts the `Got Xiaomi` blocks (sensor
+  publishes are not logged at DEBUG) and compares them with `expected.json`.
+- On ESP32 the tracker's own triggers sit in `listeners_`, which runs before `neutral_listeners_` (the
+  ble_device_base devices), so an `on_ble_service_data_advertise` never sees what a Xiaomi listener did to the
+  shared data. Check that directly: `xb_rewrite_probe(id(xa))` builds an `ESPBTDevice` with `parse_scan_rst()`,
+  calls the component's `parse_device()` and logs the service data before and after (`[probe` lines).
+- `run.sh inject pr|base` builds, flashes, captures and prints PASS when the counts match that build's model and,
+  on the PR build, the probe's service data is the same before and after `parse_device()`.
+  Base dropped a second device's frame with the same counter, the frame after a bad tag and a first frame with
+  counter 0, and rewrote the shared data in place; the PR fixed all of them.
+
+## Over the air with a second ESP32
+
+- Advertiser: an ESP32-S3 with Ethernet (OTA over Ethernet), `xbadv.yaml` with `adv_frames.hpp` from
+  `gen.py --ota`. Per frame: stop advertising, `esp_ble_gap_set_rand_addr`, `config_adv_data_raw`, start
+  non-connectable advertising with `own_addr_type` random, 20 ms interval, 250 ms dwell. Random static addresses
+  need the top two bits set, so these use C4:C1:38 instead of A4. `esp32_ble:` alone brings up the stack. The
+  web_server button starts a run: `curl -X POST -d "" http://<advertiser>/button/start/press` (no body gives 411).
+  Its `failed sends` sensor counts only GAP calls rejected when queued; a frame the controller drops later shows
+  up as missing publishes in `count.py`, the same as radio loss.
+- Receiver: `xbrecv.yaml` on the devkit, passive scan with window = interval = 100 ms. `run.sh ota pr|base` (with
+  `ADV` set) flashes it, resets it with a capture and presses start 18 s later, so the counter 0 boot frame is the
+  first one it hears.
+- Each frame arrives about 4 times; the counts matched the injection run exactly apart from one radio loss, so the
+  over the air comparison allows two missing publishes per tag.

.agents/skills/hardware-test/scripts/bk72xx_flash/partdiff.py

@@ -0,0 +1,49 @@
+"""Compare two full flash dumps of a BK72xx board per partition of its LibreTiny board layout.
+
+usage: partdiff.py BEFORE.bin AFTER.bin LAYOUT.json
+LAYOUT.json is a board base file from the LibreTiny platform (for example boards/_base/beken-72xx.json); its
+"flash" map ("name": "0xSTART+0xLENGTH") names the partitions. Prints the 4 KB sectors that changed in each one,
+plus any changed sector outside the map.
+"""
+
+import json
+import pathlib
+import sys
+
+before, after = (pathlib.Path(p).read_bytes() for p in sys.argv[1:3])
+if len(before) != len(after):
+    sys.exit(
+        f"dump sizes differ: {len(before)} and {len(after)} bytes; compare two whole chip reads"
+    )
+parts = []
+for name, span in json.loads(pathlib.Path(sys.argv[3]).read_text(encoding="utf-8"))[
+    "flash"
+].items():
+    start, length = (int(x, 16) for x in span.split("+"))
+    parts.append((start, start + length, name))
+parts.sort()
+if (end := max(e for _, e, _ in parts)) > len(before):
+    sys.exit(
+        f"the layout reaches {end:#x} but the dumps are {len(before):#x} bytes; read the whole chip"
+    )
+SECTOR = 0x1000
+covered = set()
+for start, end, name in parts:
+    changed = [
+        a
+        for a in range(start, end, SECTOR)
+        if before[a : a + SECTOR] != after[a : a + SECTOR]
+    ]
+    covered.update(range(start, end, SECTOR))
+    listed = " ".join(hex(a) for a in changed[:4]) if 0 < len(changed) <= 4 else ""
+    print(
+        f"{name:12} {len(changed):4} of {(end - start) // SECTOR:4} sectors changed {listed}"
+    )
+other = [
+    a
+    for a in range(0, len(before), SECTOR)
+    if a not in covered and before[a : a + SECTOR] != after[a : a + SECTOR]
+]
+print(
+    f"{'unmapped':12} {len(other):4} sectors changed {' '.join(hex(a) for a in other[:8])}"
+)

.agents/skills/hardware-test/scripts/bk72xx_flash/readback.sh

@@ -0,0 +1,19 @@
+#!/bin/bash
+# usage: readback.sh TAG [START LENGTH]   read flash from a BK72xx over its adapter into TAG.bin (whole chip by default)
+# env: PORT (adapter), FAMILY (ltchiptool family, default beken-7238), FLASH_BK (path to flash_bk72xx.py),
+# LT (LibreTiny penv python, default ~/.platformio/penv/.libretiny/bin/python), SIZE (whole chip bytes, default
+# 2097152). A whole 2 MB chip takes about 190 s. The read must come back at the expected size.
+T=${1:?tag}; P=${PORT:?set PORT}; FB=${FLASH_BK:?set FLASH_BK}; LT=${LT:-$HOME/.platformio/penv/.libretiny/bin/python}
+ARGS=(); [ -n "$2" ] && ARGS=(-s "$2" -l "${3:?length}")
+WANT=$(( ${3:-${SIZE:-2097152}} ))
+: > "$T.log"
+ok=
+for i in 1 2 3; do
+  echo "### try $i" >> "$T.log"
+  rm -f "$T.bin"
+  timeout 900 "$LT" "$FB" flash read -d "$P" "${ARGS[@]}" "${FAMILY:-beken-7238}" "$T.bin" >> "$T.log" 2>&1 &&
+    [ "$(wc -c < "$T.bin" | tr -d ' ')" = "$WANT" ] && ok=1 && break
+  sleep 3
+done
+[ -n "$ok" ] || { echo "FAIL: read $T (expected $WANT bytes, see $T.log)"; exit 1; }
+shasum -a 256 "$T.bin"

.agents/skills/hardware-test/scripts/ble_event_inject/cp-inject.yaml

@@ -0,0 +1,39 @@
+# Add to a copy of a bluetooth_proxy config:  packages: {cp_inject: !include cp-inject.yaml}
+esphome:
+  includes:
+    - cp_inject.hpp
+
+logger:
+  level: VERBOSE
+  logs:
+    esp32_ble_tracker: DEBUG
+    bluetooth_proxy: DEBUG
+    bluetooth_connection: DEBUG
+    api: DEBUG
+    api.service: DEBUG
+    api.connection: DEBUG
+    api.socket: DEBUG
+    scheduler: DEBUG
+    sensor: DEBUG
+    component: DEBUG
+
+button:
+  - platform: template
+    name: Inject conn param events
+    # The handlers push into a queue that takes one writer only, the Bluetooth task. Stop the scan (and keep the
+    # proxy free of real connections) so nothing else writes while the button does.
+    on_press:
+      - esp32_ble_tracker.stop_scan:
+      - wait_until:
+          condition:
+            lambda: return !esp32_ble_tracker::global_esp32_ble_tracker->scan_running();
+          timeout: 5s
+      - delay: 1s
+      - lambda: |-
+          if (esp32_ble_tracker::global_esp32_ble_tracker->scan_running()) {
+            ESP_LOGE("cpinject", "scan still running after 5 s, nothing injected");
+            return;
+          }
+          cpinject::run();
+      - esp32_ble_tracker.start_scan:
+          continuous: true

.agents/skills/hardware-test/scripts/ble_event_inject/cp_inject.hpp

@@ -0,0 +1,68 @@
+#pragma once
+// Feeds fabricated conn param update and disconnect events into ESP32BLE's static handlers.
+// The event queue takes a single writer: call run() only while scanning is stopped and no connection is active.
+#include <cstring>
+#include "esphome/components/esp32_ble/ble.h"
+#include "esphome/core/log.h"
+
+namespace cpinject {
+using GapFn = void (*)(esp_gap_ble_cb_event_t, esp_ble_gap_cb_param_t *);
+using GattcFn = void (*)(esp_gattc_cb_event_t, esp_gatt_if_t, esp_ble_gattc_cb_param_t *);
+GapFn steal_gap();
+GattcFn steal_gattc();
+// Explicit instantiation ignores access checks, which exposes the protected static handlers.
+template<GapFn F> struct GapThief {
+  friend GapFn steal_gap() { return F; }
+};
+template<GattcFn F> struct GattcThief {
+  friend GattcFn steal_gattc() { return F; }
+};
+template struct GapThief<&esphome::esp32_ble::ESP32BLE::gap_event_handler>;
+template struct GattcThief<&esphome::esp32_ble::ESP32BLE::gattc_event_handler>;
+
+static const char *const TAG = "cpinject";
+
+inline void peer(uint8_t n, uint8_t *bda) {
+  const uint8_t base[6] = {0xC0, 0xFF, 0xEE, 0x00, 0x00, 0x00};
+  memcpy(bda, base, 6);
+  bda[5] = n;
+}
+
+inline void update(uint8_t n, int status, int times) {
+  for (int i = 0; i < times; i++) {
+    esp_ble_gap_cb_param_t p{};
+    peer(n, p.update_conn_params.bda);
+    p.update_conn_params.status = (esp_bt_status_t) status;
+    p.update_conn_params.conn_int = 24;
+    p.update_conn_params.timeout = 400;
+    steal_gap()(ESP_GAP_BLE_UPDATE_CONN_PARAMS_EVT, &p);
+  }
+}
+
+inline void disconnect(uint8_t n) {
+  esp_ble_gattc_cb_param_t p{};
+  peer(n, p.disconnect.remote_bda);
+  p.disconnect.reason = ESP_GATT_CONN_TERMINATE_PEER_USER;
+  p.disconnect.conn_id = 0xEE;
+  // 0xFE matches no registered client, so the queued event is ignored by every connection.
+  steal_gattc()(ESP_GATTC_DISCONNECT_EVT, 0xFE, &p);
+}
+
+inline void run() {
+  ESP_LOGI(TAG, "STEP1 peers 1-4 fail 5x each (3 slots)");
+  for (uint8_t n = 1; n <= 4; n++)
+    update(n, 6, 5);
+  ESP_LOGI(TAG, "STEP2 peer 1 succeeds, then fails 3x");
+  update(1, 0, 1);
+  update(1, 6, 3);
+  ESP_LOGI(TAG, "STEP3 peer 2 disconnects, then fails 3x");
+  disconnect(2);
+  update(2, 6, 3);
+  ESP_LOGI(TAG, "STEP4 peer 4 fails 3x, slots still full");
+  update(4, 6, 3);
+  ESP_LOGI(TAG, "STEP5 peer 3 disconnects, peer 4 fails 3x");
+  disconnect(3);
+  update(4, 6, 3);
+  ESP_LOGI(TAG, "STEP6 done");
+}
+}  // namespace cpinject

.agents/skills/hardware-test/scripts/ble_server_security/client.yaml

@@ -0,0 +1,155 @@
+# Waveshare ESP32-S3 (any ESP32 with BLE) as the central. run.sh passes name, io, auth, idf and server (the
+# server's BLE MAC from its esp32_ble dump) with -s. ble_client never pairs by itself: the Pair button calls pair().
+esphome:
+  name: ${name}
+
+esp32:
+  board: esp32-s3-devkitc-1
+  variant: esp32s3
+  flash_size: 16MB
+  framework:
+    type: esp-idf
+    version: ${idf}
+
+logger:
+  hardware_uart: USB_SERIAL_JTAG
+  level: VERBOSE
+  logs:
+    esp32_ble_tracker: DEBUG
+    api: DEBUG
+    api.connection: DEBUG
+    api.service: DEBUG
+    api.socket: DEBUG
+    scheduler: DEBUG
+    component: DEBUG
+    wifi: DEBUG
+
+wifi:
+  ssid: !secret wifi_ssid
+  password: !secret wifi_password
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+esp32_ble:
+  io_capability: ${io}
+  auth_req_mode: ${auth}
+
+esp32_ble_tracker:
+
+ble_client:
+  - mac_address: ${server}
+    id: cli
+    on_connect:
+      - logger.log: "CLI connected"
+    on_disconnect:
+      - logger.log: "CLI disconnected"
+    on_numeric_comparison_request:
+      - logger.log:
+          format: "CLI numeric comparison %06u"
+          args: ["(unsigned) passkey"]
+      - ble_client.numeric_comparison_reply:
+          id: cli
+          accept: true
+
+text_sensor:
+  - platform: ble_client
+    ble_client_id: cli
+    id: secure_value
+    name: Secure value
+    service_uuid: "180F"
+    characteristic_uuid: "2A19"
+    update_interval: never
+    on_value:
+      - logger.log:
+          format: "CLI secure read: %s"
+          args: ["x.c_str()"]
+  - platform: ble_client
+    ble_client_id: cli
+    id: open_value
+    name: Open value
+    service_uuid: 5ec50000-0000-4000-8000-00000000f000
+    characteristic_uuid: 5ec50000-0000-4000-8000-00000000f001
+    update_interval: never
+    on_value:
+      - logger.log:
+          format: "CLI open read: %s"
+          args: ["x.c_str()"]
+
+  - platform: ble_client
+    ble_client_id: cli
+    id: mitm_value
+    name: Mitm value
+    service_uuid: 5ec50000-0000-4000-8000-00000000f000
+    characteristic_uuid: 5ec50000-0000-4000-8000-00000000f004
+    update_interval: never
+    on_value:
+      - logger.log:
+          format: "CLI mitm read: %s"
+          args: ["x.c_str()"]
+  - platform: ble_client
+    ble_client_id: cli
+    id: notify_value
+    name: Notify value
+    service_uuid: 5ec50000-0000-4000-8000-00000000f000
+    characteristic_uuid: 5ec50000-0000-4000-8000-00000000f003
+    notify: true
+    on_value:
+      - logger.log:
+          format: "CLI notification: %s"
+          args: ["x.c_str()"]
+
+button:
+  - platform: template
+    name: Subscribe
+    on_press:
+      - lambda: |-
+          // ble_client drops its service list after discovery; ESP-IDF's GATT server puts the CCCD right after
+          // the characteristic value, so write handle + 1
+          uint16_t chr = id(notify_value)->handle;
+          if (chr == 0) { ESP_LOGW("CLI", "notify characteristic handle unknown"); return; }
+          esp_ble_gattc_register_for_notify(id(cli)->get_gattc_if(), id(cli)->get_remote_bda(), chr);
+          uint8_t on[2] = {1, 0};
+          auto err = esp_ble_gattc_write_char_descr(id(cli)->get_gattc_if(), id(cli)->get_conn_id(), chr + 1, 2, on,
+                                                    ESP_GATT_WRITE_TYPE_RSP, ESP_GATT_AUTH_REQ_NONE);
+          ESP_LOGI("CLI", "CCCD write to handle %u (char %u) sent, err %d", chr + 1, chr, err);
+  - platform: template
+    name: Pair
+    on_press:
+      - lambda: 'ESP_LOGI("CLI", "pair() returned %d", (int) id(cli)->pair());'
+  - platform: template
+    name: Read secure
+    on_press:
+      - component.update: secure_value
+  - platform: template
+    name: Read mitm
+    on_press:
+      - component.update: mitm_value
+  - platform: template
+    name: Read open
+    on_press:
+      - component.update: open_value
+  - platform: template
+    name: Write secure
+    on_press:
+      - ble_client.ble_write:
+          id: cli
+          service_uuid: 5ec50000-0000-4000-8000-00000000f000
+          characteristic_uuid: 5ec50000-0000-4000-8000-00000000f002
+          value: [0x68, 0x65, 0x6C, 0x6C, 0x6F]
+  - platform: template
+    name: Disconnect
+    on_press:
+      - ble_client.disconnect: cli
+  - platform: template
+    name: Connect
+    on_press:
+      - ble_client.connect: cli
+  - platform: template
+    name: Remove bond
+    on_press:
+      - ble_client.remove_bond:
+          id: cli

.agents/skills/hardware-test/scripts/ble_server_security/drive.py

@@ -0,0 +1,60 @@
+"""Press the client's buttons in order and log every client line about BLE security.
+
+usage: drive.py CLIENT_HOST OUT STEP...   STEP: a button name, or wait:SECONDS   env: API_KEY
+"""
+
+import asyncio
+import os
+import pathlib
+import re
+import sys
+import time
+
+from aioesphomeapi import APIClient, ButtonInfo, LogLevel
+
+HOST, OUT, STEPS = sys.argv[1], sys.argv[2], sys.argv[3:]
+LOG = pathlib.Path(OUT).open("w", encoding="utf-8")  # noqa: SIM115  open for the whole run
+PATTERN = re.compile(
+    r"CLI|ble_client|ble_text_sensor|esp32_ble_client|[Aa]uth|[Ee]ncrypt|[Bb]ond|Error|status=|pair"
+)
+
+
+def note(s: str) -> None:
+    line = f"{time.strftime('%H:%M:%S')} {s}"
+    print(line)
+    LOG.write(line + "\n")
+    LOG.flush()
+
+
+async def main() -> None:
+    c = APIClient(HOST, 6053, None, noise_psk=os.environ.get("API_KEY"))
+    for attempt in range(40):
+        try:
+            await c.connect(login=True)
+            break
+        except Exception as err:  # noqa: BLE001
+            last_err = err
+            note(f"connect attempt {attempt + 1} failed: {err!r}")
+            await asyncio.sleep(2)
+    else:
+        raise last_err
+    ents, _ = await c.list_entities_services()
+    btn = {e.name: e.key for e in ents if isinstance(e, ButtonInfo)}
+
+    def on_log(m) -> None:
+        t = re.sub(r"\x1b\[[0-9;]*m", "", m.message.decode(errors="replace"))
+        if PATTERN.search(t):
+            note("LOG " + t.strip().replace("\n", " | "))
+
+    c.subscribe_logs(on_log, log_level=LogLevel.LOG_LEVEL_DEBUG)
+    for step in STEPS:
+        if step.startswith("wait:"):
+            await asyncio.sleep(float(step[5:]))
+            continue
+        note(f"PRESS {step}")
+        c.button_command(btn[step])
+        await asyncio.sleep(4)
+    await c.disconnect()
+
+
+asyncio.run(main())

.agents/skills/hardware-test/scripts/ble_server_security/run.sh

@@ -0,0 +1,32 @@
+#!/bin/bash
+# usage: run.sh TAG IO AUTH   build and flash the server (serial) and the client (OTA), then pair once and reconnect twice
+# Copy this directory into a scratch directory with a secrets.yaml (wifi_ssid, wifi_password) and run it there.
+# IO/AUTH: none bond (Just Works) or display_yes_no sc_mitm_bond (numeric comparison, MITM).
+# env: TREE (esphome checkout for the client), SERVER_TREE (checkout with esp32_ble_server security, #17562 or
+# later; default TREE), SERVER_PORT (devkit serial port), CLIENT_HOST (client address), SERVER_BLE_MAC (from the
+# server's esp32_ble dump), SC (path to serial_capture.py), IDF (default recommended), PY (default $TREE/venv/bin/python)
+# A server keeps its bonds in NVS and refuses a weaker re-pairing: erase its flash before switching from nc to jw.
+D=$(cd "$(dirname "$0")" && pwd); G=${1:?tag}; IO=${2:?io}; AUTH=${3:?auth}; T=${TREE:?set TREE}; PY=${PY:-$T/venv/bin/python}
+P=${SERVER_PORT:?set SERVER_PORT}; C=${CLIENT_HOST:?set CLIENT_HOST}; MAC=${SERVER_BLE_MAC:?set SERVER_BLE_MAC}; SC=${SC:?set SC}
+E=(-s io "$IO" -s auth "$AUTH" -s idf "${IDF:-recommended}")
+ST=${SERVER_TREE:-$T}
+build() { local t=$1; shift; (cd "$t" && PYTHONUNBUFFERED=1 timeout 1800 $PY -m esphome "$@"); }
+ok=
+for i in 1 2 3; do
+  build "$ST" -s name "bss-$G" "${E[@]}" run --no-logs "$D/server.yaml" --device "$P" > "$D/server-$G-upload.log" 2>&1 && ok=1 && break
+  sleep 5
+done
+[ -n "$ok" ] || { echo "FAIL: server build or flash (does SERVER_TREE have #17562?)"; exit 1; }
+build "$T" -s name "bsc-$G" "${E[@]}" -s server "$MAC" run --no-logs "$D/client.yaml" --device "$C" > "$D/client-$G-upload.log" 2>&1 ||
+  { echo "FAIL: client build or upload"; exit 1; }
+rm -f "$D/server-$G.txt"
+$PY "$SC" "$P" 330 "$D/server-$G.txt" --reset > /dev/null 2>&1 & CAP=$!
+trap 'kill $CAP 2>/dev/null' EXIT
+sleep 25
+R=("Disconnect" "wait:3" "Connect" "wait:10" "Pair" "wait:5" "Read secure" "Read mitm")
+$PY "$D/drive.py" "$C" "$D/client-$G.txt" "wait:6" "Remove bond" "Disconnect" "wait:3" "Connect" "wait:10" \
+  "Read secure" "Read mitm" "Write secure" "Subscribe" "wait:8" "Pair" "wait:6" "Read secure" "Read mitm" "Write secure" \
+  "Subscribe" "wait:10" "${R[@]}" "${R[@]}" > /dev/null || { echo "FAIL: driver"; exit 1; }
+wait $CAP; trap - EXIT
+MODE=jw; [ "$IO" = none ] || MODE=nc
+$PY "$D/summary.py" "$MODE" "$D/client-$G.txt" "$D/server-$G.txt"

.agents/skills/hardware-test/scripts/ble_server_security/server.yaml

@@ -0,0 +1,85 @@
+# Classic ESP32 devkit as the secure GATT server. run.sh passes name, io (none or display_yes_no),
+# auth (bond or sc_mitm_bond) and idf (recommended, or a version such as 5.5.5) with -s.
+# Needs esp32_ble_server security from #17562 (read_encrypt, numeric comparison); build it with SERVER_TREE.
+esphome:
+  name: ${name}
+
+esp32:
+  board: esp32dev
+  framework:
+    type: esp-idf
+    version: ${idf}
+
+logger:
+  level: DEBUG
+
+wifi:
+  ssid: !secret wifi_ssid
+  password: !secret wifi_password
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+esp32_ble:
+  io_capability: ${io}
+  auth_req_mode: ${auth}
+
+esp32_ble_server:
+  id: srv
+  on_connect:
+    - logger.log: "SRV client connected"
+  on_disconnect:
+    - logger.log: "SRV client disconnected"
+  on_numeric_comparison_request:
+    - logger.log:
+        format: "SRV numeric comparison %06u"
+        args: ["(unsigned) passkey"]
+    - ble_server.numeric_comparison_reply:
+        id: srv
+        address: !lambda return address;
+        accept: true
+  services:
+    - uuid: "180F"
+      characteristics:
+        - uuid: "2A19"
+          read: true
+          read_encrypt: true
+          value:
+            data: "99.9"
+    - uuid: 5ec50000-0000-4000-8000-00000000f000
+      characteristics:
+        - uuid: 5ec50000-0000-4000-8000-00000000f001
+          read: true
+          value:
+            data: "open"
+        - uuid: 5ec50000-0000-4000-8000-00000000f002
+          write: true
+          write_encrypt: true
+          on_write:
+            - lambda: 'ESP_LOGI("SRV", "secure write received: %s", std::string(x.begin(), x.end()).c_str());'
+        - uuid: 5ec50000-0000-4000-8000-00000000f004
+          read: true
+          read_encrypt_mitm: true
+          value:
+            data: "mitm"
+        - uuid: 5ec50000-0000-4000-8000-00000000f003
+          id: secure_notify
+          read: true
+          read_encrypt: true
+          notify: true
+          value:
+            data: "n0"
+
+interval:
+  - interval: 2s
+    then:
+      - lambda: |-
+          static int n = 0;
+          char buf[12];
+          snprintf(buf, sizeof(buf), "n%d", ++n);
+          id(secure_notify)->set_value((const uint8_t *) buf, strlen(buf));
+      - ble_server.characteristic.notify:
+          id: secure_notify

.agents/skills/hardware-test/scripts/ble_server_security/summary.py

@@ -0,0 +1,78 @@
+"""Judge one run.sh run from the client log (drive.py output) and the server's serial capture.
+
+usage: summary.py MODE CLIENT_LOG SERVER_CAPTURE   MODE: jw (Just Works) or nc (numeric comparison, MITM)
+Before the first Pair: the encrypted and MITM reads must each be tried and refused (an `Error reading char` reply),
+no notification may arrive and no encrypted write may reach the server. After it: the encrypted read returns 99.9,
+the write reaches the server, the second Subscribe brings notifications, and the MITM-only characteristic is readable
+only after an nc pairing. Prints the reconnect pattern and PASS or FAIL.
+"""
+
+import pathlib
+import re
+import sys
+
+MODE = sys.argv[1]
+cli = pathlib.Path(sys.argv[2]).read_text(encoding="utf-8").splitlines()
+srv = re.sub(
+    r"\x1b\[[0-9;]*m",
+    "",
+    pathlib.Path(sys.argv[3]).read_bytes().decode(errors="replace"),
+)
+pair_at = next((i for i, line in enumerate(cli) if "PRESS Pair" in line), None)
+if pair_at is None:
+    sys.exit("FAIL: no Pair step in the client log")
+before, after = cli[:pair_at], cli[pair_at:]
+subs = [i for i, line in enumerate(after) if "PRESS Subscribe" in line]
+after_sub = after[subs[0] :] if subs else []
+
+
+def count(lines: list[str], pat: str) -> int:
+    return sum(1 for line in lines if re.search(pat, line))
+
+
+def outcome(lines: list[str], button: str, ok_pat: str) -> str:
+    """'ok', 'refused' or 'none' for the first press of BUTTON.
+
+    The window runs to the next read, pair or connection step, since a reply can come after a write press. A value
+    wins over an error line: the error can belong to another characteristic read in the same window.
+    """
+    start = next((i for i, line in enumerate(lines) if f"PRESS {button}" in line), None)
+    if start is None:
+        return "none"
+    window = []
+    for line in lines[start + 1 :]:
+        if re.search(r"PRESS (Read|Pair|Disconnect|Connect|Subscribe|Remove)", line):
+            break
+        window.append(line)
+    if any(re.search(ok_pat, line) for line in window):
+        return "ok"
+    return "refused" if any("Error reading char" in line for line in window) else "none"
+
+
+secure_before = outcome(before, "Read secure", r"secure read: 9")
+mitm_before = outcome(before, "Read mitm", r"mitm read: m")
+mitm_after = outcome(after, "Read mitm", r"mitm read: m")
+checks = {
+    f"encrypted read refused before pairing ({secure_before})": secure_before
+    == "refused",
+    f"MITM read refused before pairing ({mitm_before})": mitm_before == "refused",
+    "no notifications before pairing": count(before, r"notification: n") == 0,
+    "encrypted read after pairing": outcome(after, "Read secure", r"secure read: 99\.9")
+    == "ok",
+    "encrypted write reached the server": "secure write received" in srv,
+    "notifications after subscribing paired": count(after_sub, r"notification: n") > 0,
+}
+if MODE == "nc":
+    checks[f"MITM read after nc pairing ({mitm_after})"] = mitm_after == "ok"
+else:
+    checks[f"MITM read still refused after Just Works ({mitm_after})"] = (
+        mitm_after == "refused"
+    )
+for name, ok in checks.items():
+    print(f"{'ok  ' if ok else 'FAIL'} {name}")
+print(
+    f"server: {srv.count('Authentication complete')} authentications, {srv.count('SRV numeric comparison')} numeric "
+    f"comparisons, {srv.count('Deleting unpaired device')} bonds dropped by the stack"
+)
+print("PASS" if all(checks.values()) else "FAIL")
+sys.exit(0 if all(checks.values()) else 1)

.agents/skills/hardware-test/scripts/ble_terminate/an.py

@@ -0,0 +1,90 @@
+"""Match each proxy disconnect with the peer's own disconnect line and count the links the peer kept.
+Usage: an.py DIR LABEL [ROUNDS]   reads LABEL-drive.txt, LABEL-peer.txt (serial_capture --ms) and LABEL-proxy.txt
+(esphome logs, VERBOSE). A round counts only when the last `Conn params updated` the proxy logged since the previous
+disconnect shows latency=10; otherwise it is inconclusive. Exits 1 when the peer kept a link past 3 s (the terminate
+was lost), a round was inconclusive, or fewer than ROUNDS rounds completed."""
+
+import datetime as dt
+from pathlib import Path
+import re
+import sys
+
+D, label = Path(sys.argv[1]), sys.argv[2]
+ROUNDS = int(sys.argv[3]) if len(sys.argv) > 3 else None
+
+
+def read(name: str) -> list[str]:
+    return (
+        (D / f"{label}-{name}.txt")
+        .read_text(encoding="utf-8", errors="replace")
+        .splitlines()
+    )
+
+
+def ptime(line: str) -> float | None:
+    """Seconds since midnight from a leading [HH:MM:SS.mmm] stamp."""
+    m = re.match(r"\[(\d+):(\d+):(\d+)\.(\d+)\]", line)
+    return (
+        int(m[1]) * 3600 + int(m[2]) * 60 + int(m[3]) + int(m[4]) / 1000 if m else None
+    )
+
+
+pdisc = [
+    t
+    for line in read("peer")
+    if "PEER DISCONNECT" in line and (t := ptime(line)) is not None
+]
+params = [
+    (t, int(m[1]))
+    for line in read("proxy")
+    if "Conn params updated" in line
+    and (m := re.search(r"latency=(\d+)", line))
+    and (t := ptime(line)) is not None
+]
+res = []
+prev = -1.0
+for line in read("drive"):
+    m = re.search(
+        r"R(\d+) .*DISC_CALL ([\d.]+) DISC_DONE ([\d.]+) \(\+([\d.]+)s\)", line
+    )
+    if not m:
+        if line.startswith("R"):
+            print("  ", line)
+        continue
+    # Both captures stamp local wall clock time; compare as seconds since midnight.
+    t = dt.datetime.fromtimestamp(float(m[2])).astimezone()
+    tc = t.hour * 3600 + t.minute * 60 + t.second + t.microsecond / 1e6
+    after = [p - tc for p in pdisc if -1 < p - tc < 12]
+    window = [lat for pt, lat in params if prev < pt <= tc]
+    res.append(
+        (
+            int(m[1]),
+            float(m[4]),
+            after[0] if after else None,
+            window[-1] if window else None,
+        )
+    )
+    prev = tc
+for r, done, pd, lat in res:
+    peer = (
+        f"peer disconnect +{pd:.2f}s" if pd is not None else "peer disconnect not seen"
+    )
+    print(f"  R{r:2} central done +{done:.3f}s  {peer}  latency at disconnect {lat}")
+late = sum(1 for _, _, pd, lat in res if lat == 10 and (pd is None or pd > 3))
+inconclusive = sum(1 for *_, lat in res if lat != 10)
+print(
+    f"{label}: rounds {len(res)}, peer kept the link (>3 s or not seen) {late}, inconclusive {inconclusive}"
+)
+failed = []
+if not res:
+    failed.append("no round completed")
+if ROUNDS is not None and len(res) < ROUNDS:
+    failed.append(f"only {len(res)} of {ROUNDS} rounds completed")
+if late:
+    failed.append(f"{late} links kept")
+if inconclusive:
+    failed.append(
+        f"INCONCLUSIVE: {inconclusive} rounds had no latency=10 update before the disconnect"
+    )
+print(f"PASS: {label}" if not failed else f"FAIL: {label}: " + "; ".join(failed))
+sys.exit(1 if failed else 0)

.agents/skills/hardware-test/scripts/ble_terminate/btdrive.py

@@ -0,0 +1,100 @@
+"""Connect to the peer through the proxy, read its services, idle, disconnect; print host timestamps per round.
+Usage: btdrive.py PROXY KEY PEER_MAC ROUNDS IDLE_S"""
+
+import asyncio
+import sys
+import time
+
+from aioesphomeapi import APIClient, APIConnectionError
+
+
+async def main(host: str, key: str, peer: str, rounds: int, idle: float) -> int:
+    addr = int(peer.replace(":", ""), 16)
+    cli = APIClient(host, 6053, None, noise_psk=key)
+    await cli.connect(login=True)
+    info = await cli.device_info()
+    print(f"proxy {info.name} bt_mac {info.bluetooth_mac_address}", flush=True)
+    seen = 0
+
+    def adv(msg) -> None:
+        nonlocal seen
+        seen += sum(a.address == addr for a in msg.advertisements)
+
+    # The proxy only connects to a device its scanner has seen.
+    cli.subscribe_bluetooth_le_raw_advertisements(adv)
+    await asyncio.sleep(5)
+    print(f"peer advertisements seen in 5 s: {seen}", flush=True)
+    if not seen:
+        print("FAIL: the proxy does not see the peer", flush=True)
+        await cli.disconnect()
+        return 1
+    done = 0
+    for r in range(1, rounds + 1):
+        state: asyncio.Queue = asyncio.Queue()
+
+        def cb(
+            connected: bool, mtu: int, err: int, state: asyncio.Queue = state
+        ) -> None:
+            state.put_nowait((connected, err, time.time()))
+
+        t0 = time.time()
+        try:
+            cancel = await cli.bluetooth_device_connect(
+                addr, cb, timeout=20, address_type=0, feature_flags=0xFF
+            )
+        except (APIConnectionError, TimeoutError) as e:
+            print(f"R{r} CONNECT ERROR {type(e).__name__}", flush=True)
+            await asyncio.sleep(9)
+            continue
+        try:
+            c, err, t = await asyncio.wait_for(state.get(), 25)
+        except TimeoutError:
+            print(f"R{r} CONNECT TIMEOUT", flush=True)
+            cancel()
+            await asyncio.sleep(8)
+            continue
+        if not c:
+            print(f"R{r} CONNECT FAIL err {err} after {t - t0:.2f}s", flush=True)
+            cancel()
+            await asyncio.sleep(8)
+            continue
+        try:
+            n = len((await cli.bluetooth_gatt_get_services(addr)).services)
+        except (APIConnectionError, TimeoutError) as e:
+            n = f"err {e!r}"
+        await asyncio.sleep(idle)
+        td = time.time()
+        await cli.bluetooth_device_disconnect(addr)
+        try:
+            c2, err2, t2 = await asyncio.wait_for(state.get(), 15)
+            if c2 or not isinstance(n, int):
+                # an unexpected connect state or a failed service read: the round did not run as intended
+                print(
+                    f"R{r} BAD ROUND services {n} connected {c2} err {err2}", flush=True
+                )
+            else:
+                print(
+                    f"R{r} services {n} DISC_CALL {td:.3f} DISC_DONE {t2:.3f} (+{t2 - td:.3f}s) err {err2}",
+                    flush=True,
+                )
+                done += 1
+        except TimeoutError:
+            print(
+                f"R{r} services {n} DISC_CALL {td:.3f} no disconnect state", flush=True
+            )
+        cancel()
+        await asyncio.sleep(
+            9
+        )  # a peer that missed the terminate hits its 6 s supervision timeout in this gap
+    await cli.disconnect()
+    print(f"rounds completed {done} of {rounds}", flush=True)
+    return 0 if done == rounds else 1
+
+
+sys.exit(
+    asyncio.run(
+        main(
+            sys.argv[1], sys.argv[2], sys.argv[3], int(sys.argv[4]), float(sys.argv[5])
+        )
+    )
+)

.agents/skills/hardware-test/scripts/ble_terminate/btpeer.hpp

@@ -0,0 +1,17 @@
+#pragma once
+#include <cstring>
+#include <esp_gap_ble_api.h>
+#include "esphome/core/log.h"
+#include "central_bda.hpp"  // written by run.sh from PROXY_BT_MAC
+
+// 50 ms interval, peripheral latency 10, 6 s supervision timeout
+inline void btpeer_request_latency() {
+  esp_ble_conn_update_params_t p{};
+  memcpy(p.bda, CENTRAL_BDA, 6);
+  p.min_int = 40;
+  p.max_int = 40;
+  p.latency = 10;
+  p.timeout = 600;
+  esp_err_t err = esp_ble_gap_update_conn_params(&p);
+  ESP_LOGI("peer", "requested interval 50 ms latency 10 timeout 6 s: %d", err);
+}

.agents/skills/hardware-test/scripts/ble_terminate/btpeer.yaml

@@ -0,0 +1,36 @@
+esphome:
+  name: btpeer
+  includes:
+    - btpeer.hpp
+    - central_bda.hpp
+
+esp32:
+  board: esp32dev
+  framework:
+    type: esp-idf
+
+logger:
+  level: VERBOSE
+  logs:
+    esp32_ble: VERBOSE
+    scheduler: DEBUG
+    component: DEBUG
+
+esp32_ble:
+  max_connections: 1
+
+esp32_ble_server:
+  manufacturer: test
+  on_connect:
+    - lambda: ESP_LOGI("peer", "PEER CONNECT conn %u", (unsigned) id);
+    - delay: 4s
+    - lambda: btpeer_request_latency();
+  on_disconnect:
+    - lambda: ESP_LOGI("peer", "PEER DISCONNECT conn %u", (unsigned) id);
+  services:
+    - uuid: 6e400001-b5a3-f393-e0a9-e50e24dcca9e
+      advertise: true
+      characteristics:
+        - uuid: 6e400002-b5a3-f393-e0a9-e50e24dcca9e
+          read: true
+          value: "peer"

.agents/skills/hardware-test/scripts/ble_terminate/btproxy.yaml

@@ -0,0 +1,44 @@
+substitutions:
+  name: btproxy
+
+esphome:
+  name: ${name}
+
+esp32:
+  board: esp32-s3-devkitc-1
+  variant: esp32s3
+  flash_size: 16MB
+  framework:
+    type: esp-idf
+
+logger:
+  hardware_uart: USB_SERIAL_JTAG
+  level: VERBOSE
+  logs:
+    scheduler: DEBUG
+    component: DEBUG
+    esp32_ble_tracker: DEBUG
+
+ethernet:
+  type: W5500
+  clk_pin: GPIO13
+  mosi_pin: GPIO11
+  miso_pin: GPIO12
+  cs_pin: GPIO14
+  interrupt_pin: GPIO10
+  reset_pin: GPIO9
+
+api:
+  encryption:
+    key: !secret api_key
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+esp32_ble_tracker:
+  scan_parameters:
+    active: false
+
+bluetooth_proxy:
+  active: true

.agents/skills/hardware-test/scripts/ble_terminate/run.sh

@@ -0,0 +1,31 @@
+#!/bin/bash
+# usage: run.sh LABEL ROUNDS   build both boards, then ROUNDS connect/disconnect rounds through the proxy
+# Copy this directory into a scratch directory with a secrets.yaml (api_key, base64 Noise key) and run it there.
+# env: TREE (esphome checkout to build the proxy from), PORT (peer devkit serial port), PROXY (proxy address, flashed
+# over the air; the first flash of a fresh board goes over USB), PROXY_BT_MAC (the proxy's Bluetooth MAC, printed by
+# btdrive.py as bt_mac), PEER_MAC (the peer's Bluetooth MAC, base MAC + 2 on a classic ESP32), SC (path to
+# serial_capture.py), PY (checkout python), PEER_TREE (checkout for the peer, defaults to TREE)
+D=$(cd "$(dirname "$0")" && pwd); L=${1:?label}; N=${2:-12}; T=${TREE:?set TREE}; PT=${PEER_TREE:-$T}
+P=${PORT:?set PORT}; S3=${PROXY:?set PROXY}; SC=${SC:?set SC}; PY=${PY:-$T/venv/bin/python}
+PEER=${PEER_MAC:?set PEER_MAC}; CM=${PROXY_BT_MAC:?set PROXY_BT_MAC}
+KEY=$(sed -n 's/^api_key: *"\{0,1\}\([^"]*\)"\{0,1\}$/\1/p' "$D/secrets.yaml")
+[ -n "$KEY" ] || { echo "FAIL: no api_key in secrets.yaml"; exit 1; }
+cd "$D" || exit 1
+printf '#pragma once\n#include <cstdint>\nstatic const uint8_t CENTRAL_BDA[6] = {0x%s};\n' "${CM//:/, 0x}" > central_bda.hpp
+(cd "$PT" && PYTHONUNBUFFERED=1 timeout 1500 $PY -m esphome run --no-logs "$D/btpeer.yaml" --device "$P" > "$D/$L-peer-upload.log" 2>&1) \
+  || { echo "FAIL: peer build or upload failed"; exit 1; }
+(cd "$T" && PYTHONUNBUFFERED=1 timeout 1500 $PY -m esphome run --no-logs "$D/btproxy.yaml" --device "$S3" > "$D/$L-proxy-upload.log" 2>&1) \
+  || { echo "FAIL: proxy build or upload failed"; exit 1; }
+SECS=$((N * 26 + 40))
+pkill -f "[s]erial_capture.py $P"
+$PY "$SC" "$P" $SECS "$D/$L-peer.txt" --ms > /dev/null 2>&1 & CAP=$!
+(cd "$T" && PYTHONUNBUFFERED=1 exec timeout $((SECS - 10)) $PY -m esphome logs "$D/btproxy.yaml" --device "$S3" > "$D/$L-proxy.txt" 2>&1) & LOGS=$!
+trap 'kill $CAP $LOGS 2>/dev/null' EXIT
+sleep 6
+$PY "$D/btdrive.py" "$S3" "$KEY" "$PEER" "$N" 10 > "$D/$L-drive.txt" 2>&1; RC=$?
+sleep 10
+kill $CAP $LOGS 2>/dev/null; wait 2>/dev/null; trap - EXIT
+[ -s "$D/$L-peer.txt" ] || { tail -3 "$D/$L-drive.txt"; echo "FAIL: peer capture got nothing"; exit 1; }
+$PY "$D/an.py" "$D" "$L" "$N"; ARC=$?
+[ "$RC" = 0 ] || { tail -1 "$D/$L-drive.txt"; echo "FAIL: driver rc $RC (not every round completed)"; exit 1; }
+exit $ARC

.agents/skills/hardware-test/scripts/gpio_hold/gpio-hold-esp32.yaml

@@ -0,0 +1,45 @@
+# Classic ESP32 devkit: RTC pins 25 26 27 32 33 and digital pins 19 21 22 23, three rounds 5 s after boot.
+esphome:
+  name: gpio-hold-esp32
+  includes:
+    - hold_test.hpp
+  on_boot:
+    priority: -100
+    then:
+      - delay: 5s
+      - lambda: |-
+          static const int pins[] = {25, 26, 27, 32, 33, 19, 21, 22, 23};
+          for (int r = 1; r <= 3; r++) holdtest::run(pins, sizeof(pins) / sizeof(pins[0]), r);
+
+esp32:
+  board: esp32dev
+  framework:
+    type: esp-idf
+
+logger:
+
+wifi:
+  ssid: !secret wifi_ssid
+  password: !secret wifi_password
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+# Turns on USE_GPIO_HOLD.
+output:
+  - platform: gpio
+    id: held_out
+    pin:
+      number: GPIO4
+      hold_state: true
+
+button:
+  - platform: template
+    name: rerun
+    on_press:
+      - lambda: |-
+          static const int pins[] = {25, 26, 27, 32, 33, 19, 21, 22, 23};
+          holdtest::run(pins, sizeof(pins) / sizeof(pins[0]), 9);

.agents/skills/hardware-test/scripts/gpio_hold/gpio-hold-esp32s3.yaml

@@ -0,0 +1,41 @@
+# Waveshare ESP32-S3-ETH: RTC pins 16 17 and digital pin 43 are free; a round every 15 s.
+esphome:
+  name: gpio-hold-esp32s3
+  includes:
+    - hold_test.hpp
+
+esp32:
+  board: esp32-s3-devkitc-1
+  variant: esp32s3
+  flash_size: 16MB
+  framework:
+    type: esp-idf
+
+logger:
+  hardware_uart: USB_SERIAL_JTAG
+
+wifi:
+  ssid: !secret wifi_ssid
+  password: !secret wifi_password
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+# Turns on USE_GPIO_HOLD.
+output:
+  - platform: gpio
+    id: held_out
+    pin:
+      number: GPIO44
+      hold_state: true
+
+interval:
+  - interval: 15s
+    then:
+      - lambda: |-
+          static int round = 0;
+          static const int pins[] = {16, 17, 43};
+          holdtest::run(pins, sizeof(pins) / sizeof(pins[0]), ++round);

.agents/skills/hardware-test/scripts/gpio_hold/hold_test.hpp

@@ -0,0 +1,93 @@
+#pragma once
+#include "esphome/components/esp32/gpio.h"
+#include "esphome/core/log.h"
+#include <driver/gpio.h>
+#include <driver/rtc_io.h>
+
+// Drives held pads through the ISR pin API and reads the pad back.
+namespace holdtest {
+using esphome::esp32::ESP32InternalGPIOPin;
+namespace gpio = esphome::gpio;
+static const char *const TAG = "holdtest";
+
+// One pin and one ISR argument block per GPIO and hold setting, made on first use and reused by later rounds,
+// so a test that repeats forever does not use up the heap.
+struct PinPair {
+  ESP32InternalGPIOPin *pin{nullptr};
+  esphome::ISRInternalGPIOPin isr;
+};
+static PinPair pairs[2][GPIO_NUM_MAX];
+
+static PinPair &get(int num, bool hold) {
+  PinPair &pp = pairs[hold][num];
+  if (pp.pin == nullptr) {
+    pp.pin = new ESP32InternalGPIOPin();  // NOLINT
+    pp.pin->set_pin(static_cast<gpio_num_t>(num));
+    pp.pin->set_flags(gpio::FLAG_INPUT | gpio::FLAG_OUTPUT | (hold ? gpio::FLAG_HOLD : gpio::FLAG_NONE));
+    pp.isr = pp.pin->to_isr();
+  }
+  pp.pin->setup();  // the previous round left the pad unheld with no mode
+  return pp;
+}
+
+static int lvl(int num) {
+  esphome::delayMicroseconds(50);
+  return gpio_get_level(static_cast<gpio_num_t>(num));
+}
+
+// Returns failures for one pin.
+static int run_pin(int num, bool hold) {
+  int fail = 0;
+  auto &isr = get(num, hold).isr;
+  isr.pin_mode(gpio::FLAG_INPUT | gpio::FLAG_OUTPUT);
+  int wr_ok = 0;
+  for (int i = 0; i < 20; i++) {
+    bool v = i & 1;
+    isr.digital_write(v);
+    if (lvl(num) == (int) v) {
+      wr_ok++;
+    } else {
+      fail++;
+    }
+  }
+  // Control: a raw register write without a hold refresh must not reach a held pad.
+  int before = lvl(num);
+  gpio_set_level(static_cast<gpio_num_t>(num), !before);
+  int ctrl = lvl(num);
+  bool ctrl_ok = hold ? (ctrl == before) : (ctrl != before);
+  if (!ctrl_ok)
+    fail++;
+  isr.digital_write(before);  // back in sync
+
+  // Pulls through the ISR pin_mode: input must toggle in the flags diff to apply them.
+  int pd1, pu, pd2;
+  isr.pin_mode(gpio::FLAG_NONE);
+  isr.pin_mode(gpio::FLAG_INPUT | gpio::FLAG_PULLDOWN);
+  pd1 = lvl(num);
+  isr.pin_mode(gpio::FLAG_NONE);
+  isr.pin_mode(gpio::FLAG_INPUT | gpio::FLAG_PULLUP);
+  pu = lvl(num);
+  isr.pin_mode(gpio::FLAG_NONE);
+  isr.pin_mode(gpio::FLAG_INPUT | gpio::FLAG_PULLDOWN);
+  pd2 = lvl(num);
+  if (pd1 != 0 || pu != 1 || pd2 != 0)
+    fail++;
+  ESP_LOGI(TAG, "GPIO%d hold=%d rtc=%d write %d/20 control %s pulls pd=%d pu=%d pd=%d -> %s", num, hold,
+           (int) rtc_gpio_is_valid_gpio(static_cast<gpio_num_t>(num)), wr_ok, ctrl_ok ? "ok" : "BAD", pd1, pu, pd2,
+           fail ? "FAIL" : "PASS");
+  // Leave the pad free for the next run.
+  if (hold)
+    gpio_hold_dis(static_cast<gpio_num_t>(num));
+  isr.pin_mode(gpio::FLAG_NONE);
+  return fail;
+}
+
+static void run(const int *pins, int n, int round) {
+  int fail = 0;
+  for (int i = 0; i < n; i++) {
+    fail += run_pin(pins[i], true);
+    fail += run_pin(pins[i], false);
+  }
+  ESP_LOGI(TAG, "ROUND %d %s failures=%d", round, fail ? "FAIL" : "PASS", fail);
+}
+}  // namespace holdtest

.agents/skills/hardware-test/scripts/i2c_slave/i2cbk.yaml

@@ -0,0 +1,79 @@
+# BK7238 bit banged master; needs the BK72xx software I2C bus from #21586 (dev rejects the pins).
+# The ESP32 control is the same i2c, sensor and interval blocks on a devkit (sda GPIO21, scl GPIO22, 25 or 50 kHz).
+substitutions:
+  frequency: 100kHz
+
+esphome:
+  name: i2cbk
+
+bk72xx:
+  board: xh-wb3s
+
+logger:
+  hardware_uart: UART1
+  level: DEBUG
+
+wifi:
+  ssid: !secret wifi_ssid
+  password: !secret wifi_password
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+i2c:
+  id: bus
+  sda: P26
+  scl: P24
+  scan: true
+  frequency: ${frequency}
+  timeout: 20ms  # clock stretch limit; the bus default (1 ms) is shorter than the slave's 5 ms stretch
+
+sensor:
+  - platform: hdc1080
+    address: 0x40
+    update_interval: 5s
+    temperature:
+      name: fake temperature
+    humidity:
+      name: fake humidity
+
+globals:
+  - id: n
+    type: uint8_t
+    initial_value: "0"
+  - id: n_ok
+    type: uint32_t
+    initial_value: "0"
+  - id: n_bad
+    type: uint32_t
+    initial_value: "0"
+
+interval:
+  - interval: 2s
+    then:
+      - lambda: |-
+          uint8_t k = ++id(n);
+          uint8_t w[4] = {0x20, k, (uint8_t) ~k, 0xA5};
+          uint32_t t0 = micros();
+          auto e1 = id(bus)->write(0x40, w, 4);
+          uint8_t reg = 0x20, r[3] = {0, 0, 0};
+          auto e2 = id(bus)->write_readv(0x40, &reg, 1, r, 3);
+          uint32_t t1 = micros();
+          uint8_t x;
+          auto e3 = id(bus)->read(0x41, &x, 1);
+          uint8_t s1 = 0x7E, s2 = 0x7F, rr[2];
+          uint32_t t2 = micros();
+          auto e4 = id(bus)->write_readv(0x40, &s1, 1, rr, 2);
+          uint32_t t3 = micros();
+          auto e5 = id(bus)->write_readv(0x40, &s2, 1, rr, 2);
+          uint32_t t4 = micros();
+          // the absent address must NACK and both clock stretched reads must succeed
+          bool ok = e1 == 0 && e2 == 0 && r[0] == k && r[1] == (uint8_t) ~k && r[2] == 0xA5 && e3 != 0 && e4 == 0 &&
+                    e5 == 0;
+          if (ok) id(n_ok)++; else id(n_bad)++;
+          ESP_LOGI("t", "RW %s e1 %d e2 %d got %02X %02X %02X want %02X %02X A5 | %u us | absent e %d | stretch500 e %d %u us | stretch5ms e %d %u us | ok %u bad %u",
+                   ok ? "OK" : "BAD", e1, e2, r[0], r[1], r[2], k, (uint8_t) ~k, (unsigned) (t1 - t0), e3, e4,
+                   (unsigned) (t3 - t2), e5, (unsigned) (t4 - t3), (unsigned) id(n_ok), (unsigned) id(n_bad));

.agents/skills/hardware-test/scripts/i2c_slave/i2cslave.hpp

@@ -0,0 +1,50 @@
+#pragma once
+#include <Wire.h>
+#include "esphome/core/log.h"
+
+// Fake HDC1080 at 0x40: 0x00 temperature (26.0 C), 0x01 humidity (50 %), other registers are a
+// 256 byte file; reading 0x7E holds SCL for 500 us, reading 0x7F for 5 ms.
+static volatile uint8_t SL_REGS[256];
+static volatile uint8_t sl_ptr = 0;
+static volatile uint32_t sl_rx = 0, sl_tx = 0, sl_stretch = 0;
+
+static void sl_on_receive(int n) {
+  if (n < 1)
+    return;
+  sl_ptr = Wire.read();
+  uint8_t p = sl_ptr;
+  while (--n > 0 && Wire.available())
+    SL_REGS[p++] = Wire.read();
+  sl_rx++;
+}
+
+static void sl_on_request() {
+  uint8_t b[8];
+  uint8_t p = sl_ptr;
+  sl_tx++;
+  if (p == 0x7E || p == 0x7F) {
+    sl_stretch++;
+    delayMicroseconds(p == 0x7E ? 500 : 5000);
+  }
+  for (int i = 0; i < 8; i++)
+    b[i] = SL_REGS[(uint8_t) (p + i)];
+  if (p == 0x00) {
+    b[0] = 0x66;
+    b[1] = 0x66;
+  } else if (p == 0x01) {
+    b[0] = 0x80;
+    b[1] = 0x00;
+  }
+  Wire.write(b, 8);
+}
+
+inline void i2cslave_begin() {
+  Wire.begin(4, 5, 0x40);
+  Wire.onReceive(sl_on_receive);
+  Wire.onRequest(sl_on_request);
+  ESP_LOGI("slave", "I2C slave at 0x40 on SDA GPIO4 SCL GPIO5");
+}
+
+inline void i2cslave_report() {
+  ESP_LOGI("slave", "rx %u tx %u stretch %u", (unsigned) sl_rx, (unsigned) sl_tx, (unsigned) sl_stretch);
+}

.agents/skills/hardware-test/scripts/i2c_slave/i2cslave.yaml

@@ -0,0 +1,29 @@
+esphome:
+  name: i2cslave
+  libraries:
+    - Wire
+  includes:
+    - i2cslave.hpp
+  platformio_options:
+    board_build.f_cpu: 160000000L
+  on_boot:
+    priority: -100
+    then:
+      - lambda: i2cslave_begin();
+
+esp8266:
+  board: d1_mini
+
+logger:
+
+wifi:
+  ssid: !secret wifi_ssid
+  password: !secret wifi_password
+
+ota:
+  - platform: esphome
+
+interval:
+  - interval: 10s
+    then:
+      - lambda: i2cslave_report();

.agents/skills/hardware-test/scripts/ld2450/holddrive.py

@@ -0,0 +1,127 @@
+"""Time the presence hold on stream-harness.yaml and check presence_timeout survives a restart.
+
+usage: holddrive.py HOST OUT   env: API_KEY (when the harness has one)
+For timeouts 5 and 0: target frames for 3 s, then empty frames; prints the ms from the "mode none" press to the
+has_target off line (up to one 100 ms frame of that is the harness). Then sets 0 and 20, restarts, reads the number.
+A missing press line, off line or number state prints NO-DATA and the run exits 1.
+"""
+
+import asyncio
+import os
+import pathlib
+import re
+import sys
+
+from aioesphomeapi import APIClient, ButtonInfo, LogLevel, NumberInfo, NumberState
+
+HOST, OUT = sys.argv[1], sys.argv[2]
+LOG = pathlib.Path(OUT).open("w", encoding="utf-8")  # noqa: SIM115  open for the whole run
+MISSING = []
+
+
+def note(s: str) -> None:
+    print(s)
+    LOG.write(s + "\n")
+    LOG.flush()
+
+
+async def session():
+    c = APIClient(HOST, 6053, None, noise_psk=os.environ.get("API_KEY"))
+    for attempt in range(60):
+        try:
+            await c.connect(login=True)
+            break
+        except Exception as err:  # noqa: BLE001
+            last_err = err
+            note(f"connect attempt {attempt + 1} failed: {err!r}")
+            await asyncio.sleep(2)
+    else:
+        raise last_err
+    ents, _ = await c.list_entities_services()
+    btn = {e.name: e.key for e in ents if isinstance(e, ButtonInfo)}
+    num = next(
+        e.key
+        for e in ents
+        if isinstance(e, NumberInfo) and e.name == "Presence Timeout"
+    )
+    st, lines = {}, []
+
+    def on_state(s) -> None:
+        if isinstance(s, NumberState) and s.key == num:
+            st["t"] = s.state
+
+    def on_log(m) -> None:
+        t = re.sub(r"\x1b\[[0-9;]*m", "", m.message.decode(errors="replace"))
+        if "harness" in t:
+            lines.append(t)
+            LOG.write("LOG " + t + "\n")
+
+    c.subscribe_states(on_state)
+    c.subscribe_logs(on_log, log_level=LogLevel.LOG_LEVEL_DEBUG)
+    for _ in range(20):
+        await asyncio.sleep(1)
+        if "t" in st:
+            break
+    # a freshly booted ESP8266 delays API commands that arrive too early
+    await asyncio.sleep(8)
+    return c, btn, num, st, lines
+
+
+def first_ms(lines: list[str], pat: str) -> int | None:
+    for line in lines:
+        if m := re.search(pat + r" at (\d+)", line):
+            return int(m.group(1))
+    return None
+
+
+def number(st: dict, what: str) -> str:
+    if "t" not in st:
+        MISSING.append(what)
+        return "NO-DATA"
+    return f"{st['t']:g}"
+
+
+async def hold(c, btn, num, st, lines, timeout: int) -> None:
+    c.number_command(num, timeout)
+    await asyncio.sleep(2)
+    if st.get("t") != timeout:
+        # dev snaps 0 back to the saved value; the hold is then timed with what the number reads
+        note(f"HOLD timeout={timeout}: the number reads {st.get('t')}, timing that")
+    c.button_command(btn["Mode target"])
+    await asyncio.sleep(3)
+    lines.clear()  # only this run's press and off lines count
+    c.button_command(btn["Mode none"])
+    await asyncio.sleep(timeout + 8)
+    start, off = first_ms(lines, "mode none"), first_ms(lines, "presence raw OFF")
+    if start is None or off is None or off < start:
+        MISSING.append(f"hold {timeout}")
+        note(f"HOLD timeout={timeout}: NO-DATA (press line {start}, off line {off})")
+    else:
+        note(
+            f"HOLD timeout={timeout}: off {off - start} ms after the last target frame"
+        )
+
+
+async def main() -> None:
+    c, btn, num, st, lines = await session()
+    note(f"START number {number(st, 'start')}")
+    for t in (5, 0):
+        await hold(c, btn, num, st, lines, t)
+    for t in (0, 20):
+        st.pop("t", None)
+        c.number_command(num, t)
+        await asyncio.sleep(2)
+        note(f"SET {t}, number reads {number(st, f'set {t}')}")
+        c.button_command(btn["Restart"])
+        await asyncio.sleep(1)
+        await c.disconnect()
+        await asyncio.sleep(15)
+        c, btn, num, st, lines = await session()
+        note(f"AFTER RESTART number reads {number(st, f'restart {t}')}")
+    await c.disconnect()
+    if MISSING:
+        note(f"NO-DATA for: {', '.join(MISSING)}")
+        sys.exit(1)
+
+
+asyncio.run(main())

.agents/skills/hardware-test/scripts/ld2450/linkdrive.py

@@ -0,0 +1,78 @@
+"""Step stream-harness.yaml through frames, silence, garbage and radar restarts; print every state change.
+
+usage: linkdrive.py HOST OUT   env: API_KEY (when the harness has one)
+Lines are seconds since start: PRESS steps, STATE changes of the ld2450 entities (unknown when missing), and LOG lines
+about the link or ld2450 warnings.
+"""
+
+import asyncio
+import math
+import os
+import pathlib
+import re
+import sys
+import time
+
+from aioesphomeapi import APIClient, ButtonInfo, LogLevel
+
+HOST, OUT = sys.argv[1], sys.argv[2]
+LOG = pathlib.Path(OUT).open("w", encoding="utf-8")  # noqa: SIM115  open for the whole run
+T0 = time.time()
+STEPS = [
+    ("Mode none", 8), ("Mode target", 6), ("Mode silent", 10), ("Mode none", 4), ("Mode target", 4),
+    ("Mode garbage", 10), ("Mode target", 5),
+    ("Radar Restart", 0.3), ("Mode silent", 3.5), ("Mode target", 6),
+    ("Radar Restart", 0.3), ("Mode silent", 15), ("Mode none", 4),
+]  # fmt: skip
+
+
+def note(s: str) -> None:
+    line = f"{time.time() - T0:7.2f} {s}"
+    print(line)
+    LOG.write(line + "\n")
+    LOG.flush()
+
+
+async def main() -> None:
+    c = APIClient(HOST, 6053, None, noise_psk=os.environ.get("API_KEY"))
+    await c.connect(login=True)
+    ents, _ = await c.list_entities_services()
+    names = {e.key: e.name for e in ents}
+    btn = {e.name: e.key for e in ents if isinstance(e, ButtonInfo)}
+    last = {}
+
+    def on_state(s) -> None:
+        n = names.get(s.key)
+        if n is None or n in btn:
+            return
+        v = (
+            "unknown"
+            if getattr(s, "missing_state", False)
+            else getattr(s, "state", None)
+        )
+        if isinstance(v, float):
+            v = "unknown (nan)" if math.isnan(v) else round(v, 1)
+        if last.get(n) != v:
+            last[n] = v
+            note(f"STATE {n} = {v}")
+
+    def on_log(m) -> None:
+        t = re.sub(r"\x1b\[[0-9;]*m", "", m.message.decode(errors="replace"))
+        if re.search(r"[Ll]ink|\[W\]\[ld2450", t):
+            note("LOG " + t.strip())
+
+    c.subscribe_states(on_state)
+    c.subscribe_logs(on_log, log_level=LogLevel.LOG_LEVEL_DEBUG)
+    await asyncio.sleep(3)
+    for name, wait in STEPS:
+        note(f"PRESS {name}")
+        c.button_command(btn[name])
+        await asyncio.sleep(wait)
+    await c.disconnect()
+    # dev freezes entities during silence but still updates them on target frames; nothing at all is a dead stream
+    if not last:
+        note("NO-DATA: no entity state arrived")
+        sys.exit(1)
+
+
+asyncio.run(main())

.agents/skills/hardware-test/scripts/ld2450/macdrive.py

@@ -0,0 +1,82 @@
+"""Read the LD24xx BT MAC text sensors, turn every LD24xx Bluetooth switch off, then on again.
+
+usage: macdrive.py HOST OUT   env: API_KEY (when the device has one)
+Works on any device whose MAC, Bluetooth and firmware entities have those words in their names (the BT MAC sensors
+may be disabled by default; they still publish over the API). Quote at most the first MAC byte in a PR.
+"""
+
+import asyncio
+import os
+import pathlib
+import sys
+import time
+
+from aioesphomeapi import APIClient, SwitchInfo, TextSensorInfo
+
+HOST, OUT = sys.argv[1], sys.argv[2]
+LOG = pathlib.Path(OUT).open("w", encoding="utf-8")  # noqa: SIM115  open for the whole run
+
+
+def note(s: str) -> None:
+    line = f"{time.strftime('%H:%M:%S')} {s}"
+    print(line)
+    LOG.write(line + "\n")
+    LOG.flush()
+
+
+async def main() -> None:
+    c = APIClient(HOST, 6053, None, noise_psk=os.environ.get("API_KEY"))
+    await c.connect(login=True)
+    ents, _ = await c.list_entities_services()
+    keys = {
+        e.key: e.name
+        for e in ents
+        if isinstance(e, (SwitchInfo, TextSensorInfo))
+        and any(w in e.name for w in ("MAC", "Bluetooth", "Firmware"))
+    }
+    states = {}
+
+    def on_state(s) -> None:
+        if s.key in keys and states.get(keys[s.key]) != getattr(s, "state", None):
+            states[keys[s.key]] = s.state
+            note(f"STATE {keys[s.key]} = {s.state}")
+
+    c.subscribe_states(on_state)
+    await asyncio.sleep(8)
+    switches = {
+        e.name: e.key
+        for e in ents
+        if isinstance(e, SwitchInfo) and "Bluetooth" in e.name
+    }
+    macs = [n for n in keys.values() if "MAC" in n]
+    if not macs or not switches:
+        note(
+            f"FAIL: need a MAC text sensor and a Bluetooth switch, found {sorted(keys.values())}"
+        )
+        sys.exit(1)
+    missing = []
+    if silent := [n for n in sorted(keys.values()) if n not in states]:
+        note(f"NO-DATA after 8 s: {', '.join(silent)}")
+        missing += silent
+    note(
+        "SNAP start: "
+        + "; ".join(f"{n}={states.get(n)}" for n in sorted(keys.values()))
+    )
+    for phase, on in (("off", False), ("on", True)):
+        for name, key in switches.items():
+            note(f"SET {name} {phase}")
+            c.switch_command(key, on)
+            await asyncio.sleep(1)
+        await asyncio.sleep(20)
+        note(
+            f"SNAP bluetooth {phase}: "
+            + "; ".join(f"{n}={states.get(n)}" for n in sorted(keys.values()))
+        )
+    missing += [n for n in switches if states.get(n) is not True]
+    await c.disconnect()
+    if missing:
+        note(f"NO-DATA for: {', '.join(sorted(set(missing)))}")
+        sys.exit(1)
+
+
+asyncio.run(main())

.agents/skills/hardware-test/scripts/ld2450/stream-connected.yaml

@@ -0,0 +1,5 @@
+binary_sensor:
+  - platform: ld2450
+    ld2450_id: radar
+    connected:
+      name: Connected

.agents/skills/hardware-test/scripts/ld2450/stream-harness.yaml

@@ -0,0 +1,128 @@
+# D1 mini, ld2450 on UART0 looped back onto itself, fed one periodic frame every 100 ms from a mode global:
+# 0 no target, 1 target 1 approaching, 2 silent, 3 garbage bytes (what a radar at another baud rate looks like).
+# Add `connected: !include stream-connected.yaml` under packages on builds whose ld2450 has the link watchdog.
+substitutions:
+  name: ld2450-stream
+
+esphome:
+  name: ${name}
+  on_boot:
+    priority: -100
+    then:
+      - lambda: |-
+          // UART0 TX looped back into its own RX so injected frames reach the hub
+          USC0(0) |= (1 << UCLBE);
+          ESP_LOGI("harness", "UART0 loopback on");
+
+esp8266:
+  board: d1_mini
+
+wifi:
+  ssid: !secret wifi_ssid
+  password: !secret wifi_password
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+logger:
+  baud_rate: 0
+  level: DEBUG
+
+uart:
+  id: radar_uart
+  tx_pin: GPIO1
+  rx_pin: GPIO3
+  baud_rate: 256000
+
+ld2450:
+  id: radar
+  uart_id: radar_uart
+
+globals:
+  - id: mode
+    type: int
+    initial_value: "0"
+
+interval:
+  - interval: 100ms
+    then:
+      - if:
+          condition:
+            lambda: return id(mode) == 1;
+          then:
+            - uart.write:
+                id: radar_uart
+                data: [0xAA, 0xFF, 0x03, 0x00, 0x00, 0x00, 0xE8, 0x83, 0x14, 0x00, 0x64, 0x00,
+                       0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+                       0x00, 0x00, 0x00, 0x00, 0x55, 0xCC]
+      - if:
+          condition:
+            lambda: return id(mode) == 0;
+          then:
+            - uart.write:
+                id: radar_uart
+                data: [0xAA, 0xFF, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+                       0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+                       0x00, 0x00, 0x00, 0x00, 0x55, 0xCC]
+      - if:
+          condition:
+            lambda: return id(mode) == 3;
+          then:
+            - uart.write:
+                id: radar_uart
+                data: [0x3F, 0x80, 0xF8, 0x00, 0x7E, 0x07, 0xC0, 0x1F, 0xE0, 0x03, 0xF0, 0x0F]
+
+binary_sensor:
+  - platform: ld2450
+    has_target:
+      name: Has Target
+      # no settle filter, so the hub's own presence hold is what gets timed
+      filters: []
+      on_state:
+        - lambda: 'ESP_LOGI("harness", "presence raw %s at %u", x ? "ON" : "OFF", (unsigned) millis());'
+
+sensor:
+  - platform: ld2450
+    target_count:
+      name: Target Count
+    target_1:
+      x:
+        name: T1 X
+
+text_sensor:
+  - platform: ld2450
+    target_1:
+      direction:
+        name: T1 Direction
+
+number:
+  - platform: ld2450
+    presence_timeout:
+      name: Presence Timeout
+
+button:
+  - platform: template
+    name: Mode none
+    on_press:
+      - lambda: 'id(mode) = 0; ESP_LOGI("harness", "mode none at %u", (unsigned) millis());'
+  - platform: template
+    name: Mode target
+    on_press:
+      - lambda: 'id(mode) = 1; ESP_LOGI("harness", "mode target at %u", (unsigned) millis());'
+  - platform: template
+    name: Mode silent
+    on_press:
+      - lambda: 'id(mode) = 2; ESP_LOGI("harness", "mode silent at %u", (unsigned) millis());'
+  - platform: template
+    name: Mode garbage
+    on_press:
+      - lambda: 'id(mode) = 3; ESP_LOGI("harness", "mode garbage at %u", (unsigned) millis());'
+  - platform: ld2450
+    ld2450_id: radar
+    restart:
+      name: Radar Restart
+  - platform: restart
+    name: Restart

.agents/skills/hardware-test/scripts/libretiny_crash/rtl_drive.py

@@ -0,0 +1,124 @@
+"""Press one crash button after another on an RTL8720C test build and keep the decoded API logs of each next boot.
+Usage: rtl_drive.py HOST OUTDIR TREE YAML BUTTON...   (button names from templates/rtl8720c-crash-buttons-dut.yaml)
+Prints PRESS lines with the wall clock time, so resets can be timed against a serial capture taken with --ms.
+Each case waits for the board to answer on the API port again, then saves 15 s of `esphome logs` (run from TREE,
+so the CLI decodes the addresses) as OUTDIR/caseN-BUTTON.txt. The run prints CASE N BUTTON and exits 1 with NO RESET
+when the API port never closes after the press (the button did not reset the board), NO BOOT when the board is not
+back within BOOT_DEADLINE_S, and NO LOGS when the log capture never connected."""
+
+import asyncio
+import contextlib
+import os
+from pathlib import Path
+import socket
+import subprocess
+import sys
+import time
+
+from aioesphomeapi import APIClient
+
+host, out, tree, yaml = sys.argv[1:5]
+buttons = sys.argv[5:]
+BOOT_DEADLINE_S = 120
+# Every case button resets the board: a fault, the watchdog (hang) or a restart. The watchdog takes about 11 s.
+RESET_DEADLINE_S = 20
+
+
+def port_open() -> bool:
+    try:
+        socket.create_connection((host, 6053), 1).close()
+    except OSError:
+        return False
+    return True
+
+
+def wait_for_port() -> bool:
+    deadline = time.monotonic() + BOOT_DEADLINE_S
+    while not port_open():
+        if time.monotonic() > deadline:
+            return False
+        time.sleep(1)
+    return True
+
+
+def wait_for_reset() -> bool:
+    deadline = time.monotonic() + RESET_DEADLINE_S
+    while port_open():
+        if time.monotonic() > deadline:
+            return False
+        time.sleep(0.2)
+    return True
+
+
+def fail(i: int, name: str, why: str) -> None:
+    print(f"CASE {i} {name} {why}", flush=True)
+    sys.exit(1)
+
+
+async def button_keys() -> dict[str, int]:
+    cli = APIClient(host, 6053, None)
+    await cli.connect(login=True)
+    try:
+        entities, _ = await cli.list_entities_services()
+    finally:
+        await cli.disconnect()
+    return {e.name: e.key for e in entities if type(e).__name__ == "ButtonInfo"}
+
+
+async def press(name: str, key: int) -> None:
+    cli = APIClient(host, 6053, None)
+    await cli.connect(login=True)
+    now = time.time()
+    print(
+        f"PRESS {name} at {time.strftime('%H:%M:%S', time.localtime(now))}.{int(now * 1000) % 1000:03d}",
+        flush=True,
+    )
+    cli.button_command(key)
+    await asyncio.sleep(0.5)
+    # the board resets within about a second, so the disconnect reply may never come
+    with contextlib.suppress(Exception):
+        await asyncio.wait_for(cli.disconnect(), 2)
+
+
+if not wait_for_port():
+    sys.exit("NO BOOT before the first case")
+# Check every name up front, so a typo in the last case does not waste the crashes before it
+keys = asyncio.run(button_keys())
+if missing := [b for b in buttons if b not in keys]:
+    sys.exit(f"Unknown buttons {missing}; the board has {sorted(keys)}")
+
+for i, name in enumerate(buttons):
+    if not wait_for_port():
+        fail(i, name, "NO BOOT")
+    time.sleep(5)
+    asyncio.run(press(name, keys[name]))
+    if not wait_for_reset():
+        fail(i, name, "NO RESET")
+    time.sleep(20)  # boot and Wi-Fi
+    if not wait_for_port():
+        fail(i, name, "NO BOOT")
+    case_file = Path(out) / f"case{i}-{name.replace(' ', '_')}.txt"
+    with case_file.open("w", encoding="utf-8") as fh:
+        subprocess.run(
+            [
+                "timeout",
+                "15",
+                sys.executable,
+                "-m",
+                "esphome",
+                "logs",
+                yaml,
+                "--device",
+                host,
+            ],
+            cwd=tree,
+            env={**os.environ, "PYTHONPATH": tree, "PYTHONUNBUFFERED": "1"},
+            stdout=fh,
+            stderr=subprocess.STDOUT,
+            check=False,
+        )
+    if "Successful handshake" not in case_file.read_text(
+        encoding="utf-8", errors="replace"
+    ):
+        fail(i, name, "NO LOGS")
+    print(f"CASE {i} {name} done", flush=True)

.agents/skills/hardware-test/scripts/libretiny_crash/rtl_rom_vs_report.py

@@ -0,0 +1,81 @@
+"""Match each RTL8720C ROM fault dump in a serial capture with the report on the next boot.
+
+usage: rtl_rom_vs_report.py SERIAL_CAPTURE   (taken with serial_capture.py --ms)
+Per ROM dump: the ROM's CFSR and PC, ms from the dump to the next `ROM Version` banner, and the Reason, CFSR, PC and SP
+the next boot reported. The ROM lines start with a bare CR, so the capture is split on newlines only.
+"""
+
+import pathlib
+import re
+import sys
+
+RAW = pathlib.Path(sys.argv[1]).read_bytes().decode(errors="replace")
+LINES = [
+    re.sub(r"\x1b\[[0-9;]*m", "", line) for line in RAW.replace("\r", "").split("\n")
+]
+
+
+def stamp(line: str) -> float | None:
+    """Seconds since midnight from a serial_capture.py --ms timestamp."""
+    m = re.search(r"\[(\d\d):(\d\d):(\d\d\.\d{3})\]", line)
+    return (
+        int(m.group(1)) * 3600 + int(m.group(2)) * 60 + float(m.group(3)) if m else None
+    )
+
+
+if not any(stamp(line) is not None for line in LINES):
+    sys.exit(
+        "NO-DATA: no [HH:MM:SS.mmm] timestamps; take the capture with serial_capture.py --ms"
+    )
+dumps = bad = 0
+i = 0
+while i < len(LINES):
+    if "Fault Handler:" not in LINES[i]:
+        i += 1
+        continue
+    t0 = stamp(LINES[i])
+    kind = LINES[i + 2].split("]", 1)[-1].strip() if i + 2 < len(LINES) else "?"
+    cfsr = pc = None
+    j = i
+    while j < len(LINES) and "ROM Version" not in LINES[j]:
+        if cfsr is None and "Configurable Fault Status Reg" in LINES[j]:
+            cfsr = LINES[j].split("=")[1].strip()
+        if pc is None and "PC  =" in LINES[j]:
+            pc = LINES[j].split("=")[1].strip()
+        j += 1
+    tj = next(
+        (
+            stamp(LINES[x])
+            for x in range(j, min(j + 5, len(LINES)))
+            if stamp(LINES[x]) is not None
+        ),
+        None,
+    )
+    reset = (
+        f"{(tj - t0) * 1000:.0f} ms"
+        if tj is not None and t0 is not None
+        else "no reset seen"
+    )
+    report = {}
+    k = j + 1
+    while (
+        k < len(LINES)
+        and "Fault Handler:" not in LINES[k]
+        and "boot reason" not in LINES[k]
+    ):
+        m = re.search(r"(Reason|CFSR|PC|SP):\s+(.*)", LINES[k])
+        if m and "libretiny.crash" in LINES[k]:
+            report.setdefault(m.group(1), m.group(2).split("  (")[0].strip())
+        k += 1
+    dumps += 1
+    missing = [
+        n for n, v in (("ROM CFSR", cfsr), ("ROM PC", pc), ("reset", tj)) if v is None
+    ]
+    missing += [k for k in ("Reason", "CFSR", "PC", "SP") if k not in report]
+    bad += bool(missing)
+    flag = f" | MISSING {', '.join(missing)}" if missing else ""
+    print(f"{kind:18} ROM CFSR {cfsr} PC {pc} | reset {reset} | report {report}{flag}")
+    i = j
+if not dumps:
+    sys.exit("NO-DATA: no ROM fault dump in the capture")
+sys.exit(1 if bad else 0)

.agents/skills/hardware-test/scripts/light_sleep_wake/wake-probe-esp32c3.yaml

@@ -0,0 +1,40 @@
+# Any ESP32-C3 with GPIO10 free; drop ESPHome ledc outputs, the C3 shares one LEDC clock across timers.
+esphome:
+  name: wake-probe-esp32c3
+  includes:
+    - wake_probe.hpp
+  on_boot:
+    priority: -100
+    then:
+      - lambda: wakeprobe::start();
+
+esp32:
+  board: esp32-c3-devkitm-1
+  variant: esp32c3
+  framework:
+    type: esp-idf
+    advanced:
+      include_builtin_idf_components:
+        - esp_driver_ledc
+
+logger:
+
+wifi:
+  ssid: !secret wifi_ssid
+  password: !secret wifi_password
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+power_management:
+  - platform: esp32_power_management
+    enable_light_sleep: true
+    profiling: true
+
+interval:
+  - interval: 30s
+    then:
+      - lambda: wakeprobe::report();

.agents/skills/hardware-test/scripts/light_sleep_wake/wake_probe.hpp

@@ -0,0 +1,110 @@
+#pragma once
+// LEDC on an RC_FAST clock keeps toggling GPIO10 in light sleep; an ESP32InternalGPIOPin interrupt on the same pad
+// timestamps every edge.
+#include <cstdio>
+#include "driver/ledc.h"
+#include "driver/gpio.h"
+#include "esp_pm.h"
+#include "esp_timer.h"
+#include "esphome/components/esp32/gpio.h"
+#include "esphome/core/log.h"
+
+namespace wakeprobe {
+static const char *const TAG = "wakeprobe";
+static constexpr gpio_num_t PIN = GPIO_NUM_10;
+static volatile uint32_t edges = 0;
+static volatile int64_t last_us = 0;
+static volatile int64_t gap_min = INT64_MAX, gap_max = 0;
+static volatile uint32_t late_edges = 0;  // gap more than 20 ms off the half period
+static int64_t start_us = 0;
+static uint32_t half_period_us = 0;
+static uint32_t freq = 0;
+static esp_err_t err_timer = ESP_OK, err_channel = ESP_OK;
+
+static void IRAM_ATTR on_edge(void *) {
+  int64_t now = esp_timer_get_time();
+  if (last_us != 0) {
+    int64_t gap = now - last_us;
+    if (gap < gap_min)
+      gap_min = gap;
+    if (gap > gap_max)
+      gap_max = gap;
+    int64_t off = gap - (int64_t) half_period_us;
+    if (off > 20000 || off < -20000)
+      late_edges = late_edges + 1;
+  }
+  last_us = now;
+  edges = edges + 1;
+}
+
+inline void start() {
+  auto *pin = new esphome::esp32::ESP32InternalGPIOPin();  // NOLINT
+  pin->set_pin(PIN);
+  pin->set_flags(esphome::gpio::FLAG_INPUT);
+  pin->setup();
+  pin->attach_interrupt(on_edge, (void *) nullptr, esphome::gpio::INTERRUPT_ANY_EDGE);
+  ledc_timer_config_t t{};
+  t.speed_mode = LEDC_LOW_SPEED_MODE;
+  t.duty_resolution = LEDC_TIMER_14_BIT;
+  t.timer_num = LEDC_TIMER_0;
+  t.freq_hz = 2;
+  t.clk_cfg = LEDC_USE_RC_FAST_CLK;
+  esp_err_t e1 = err_timer = ledc_timer_config(&t);
+  ledc_channel_config_t c{};
+  c.gpio_num = PIN;
+  c.speed_mode = LEDC_LOW_SPEED_MODE;
+  c.channel = LEDC_CHANNEL_0;
+  c.timer_sel = LEDC_TIMER_0;
+  c.duty = 1 << 13;
+  c.sleep_mode = LEDC_SLEEP_MODE_KEEP_ALIVE;
+  esp_err_t e2 = err_channel = ledc_channel_config(&c);
+  // ledc_channel_config makes the pad output only; turn its input back on so the interrupt sees the wave.
+  gpio_input_enable(PIN);
+  freq = ledc_get_freq(LEDC_LOW_SPEED_MODE, LEDC_TIMER_0);
+  half_period_us = freq ? 500000 / freq : 0;
+  start_us = esp_timer_get_time();
+  ESP_LOGI(TAG, "LEDC timer=%s channel=%s freq=%u Hz half period %u us", esp_err_to_name(e1), esp_err_to_name(e2),
+           (unsigned) freq, (unsigned) half_period_us);
+}
+
+inline void report() {
+  if (start_us == 0)
+    return;
+  char lv[21];
+  for (int i = 0; i < 20; i++) {
+    lv[i] = gpio_get_level(PIN) ? '1' : '0';
+    vTaskDelay(pdMS_TO_TICKS(50));
+  }
+  lv[20] = 0;
+  // Elapsed time and edge count are read together, after the 1 s level loop, so they describe the same interval.
+  float secs = (esp_timer_get_time() - start_us) / 1e6f;
+  uint32_t seen = edges;
+  ESP_LOGI(TAG, "LEDC timer=%s channel=%s freq=%u, level over 1 s: %s", esp_err_to_name(err_timer),
+           esp_err_to_name(err_channel), (unsigned) freq, lv);
+  ESP_LOGI(TAG, "%.0f s: edges=%u expected=%.0f gap min=%lld max=%lld us, off by >20 ms: %u", secs, (unsigned) seen,
+           secs * 2 * freq, (long long) gap_min, (long long) gap_max, (unsigned) late_edges);
+#ifdef CONFIG_PM_PROFILING
+  static char buf[1024];
+  FILE *f = fmemopen(buf, sizeof(buf) - 1, "w");
+  if (f == nullptr) {
+    ESP_LOGW(TAG, "pm dump failed: fmemopen");
+    return;
+  }
+  esp_pm_dump_locks(f);
+  long n = ftell(f);
+  fclose(f);
+  buf[n < 0 ? 0 : n] = 0;
+  bool sleep_seen = false;
+  for (char *line = strtok(buf, "\n"); line != nullptr; line = strtok(nullptr, "\n")) {
+    if (strstr(line, "SLEEP") || strstr(line, "CPU_MAX") || strstr(line, "APB_MAX") || strstr(line, "APB_MIN")) {
+      sleep_seen = sleep_seen || strstr(line, "SLEEP") != nullptr;
+      ESP_LOGI(TAG, "pm %s", line);
+    }
+  }
+  if (!sleep_seen)
+    ESP_LOGW(TAG, "pm dump has no SLEEP line (%ld bytes, buffer %u)", n, (unsigned) sizeof(buf));
+#else
+  ESP_LOGW(TAG, "pm profiling unavailable: set power_management profiling: true");
+#endif
+}
+}  // namespace wakeprobe

.agents/skills/hardware-test/scripts/manual_ip/ap.yaml

@@ -0,0 +1,36 @@
+# AP manual IP: the only network is missing, so the fallback AP starts after 10 s on 10.42.7.1.
+substitutions:
+  name: mip-ap
+  plat: esp32
+
+esphome:
+  name: ${name}
+
+packages:
+  plat: !include plat-${plat}.yaml
+
+logger:
+  level: VERBOSE
+  logs:
+    scheduler: DEBUG
+    component: DEBUG
+
+wifi:
+  networks:
+    - ssid: mip-nowhere
+      password: nowhere123
+      manual_ip:
+        static_ip: 10.9.9.9
+        gateway: 10.9.9.1
+        subnet: 255.255.255.0
+        dns1: 10.9.9.1
+  ap:
+    ssid: mip-ap-test
+    password: mipaptest1
+    ap_timeout: 10s
+    manual_ip:
+      static_ip: 10.42.7.1
+      gateway: 10.42.7.1
+      subnet: 255.255.255.0
+
+captive_portal:

.agents/skills/hardware-test/scripts/manual_ip/client.yaml

@@ -0,0 +1,36 @@
+# Joins the DUT's AP and fetches its page every 10 s.
+substitutions:
+  name: mip-client
+  plat: esp32
+
+esphome:
+  name: ${name}
+
+packages:
+  plat: !include plat-${plat}.yaml
+
+logger:
+
+wifi:
+  ssid: mip-ap-test
+  password: mipaptest1
+  reboot_timeout: 0s
+
+http_request:
+  verify_ssl: false
+
+interval:
+  - interval: 10s
+    then:
+      - if:
+          condition: wifi.connected
+          then:
+            - http_request.get:
+                url: http://10.42.7.1/
+                capture_response: false
+                on_response:
+                  then:
+                    - lambda: ESP_LOGI("client", "GET http://10.42.7.1/ status=%d", response->status_code);
+                on_error:
+                  then:
+                    - logger.log: "GET http://10.42.7.1/ failed"

.agents/skills/hardware-test/scripts/manual_ip/dnsprobe.hpp

@@ -0,0 +1,32 @@
+#pragma once
+#include "esphome/core/log.h"
+#include "lwip/dns.h"
+#include "lwip/ip_addr.h"
+#ifdef USE_ESP8266
+#include <ESP8266WiFi.h>
+#else
+#include "lwip/netdb.h"
+#include "lwip/inet.h"
+#endif
+inline void dns_probe() {
+  char b0[IPADDR_STRLEN_MAX], b1[IPADDR_STRLEN_MAX], res[16] = "-";
+  ipaddr_ntoa_r(dns_getserver(0), b0, sizeof(b0));
+  ipaddr_ntoa_r(dns_getserver(1), b1, sizeof(b1));
+  int ok;
+#ifdef USE_ESP8266
+  IPAddress ip;
+  ok = WiFi.hostByName("pool.ntp.org", ip);
+  if (ok)
+    strncpy(res, ip.toString().c_str(), sizeof(res) - 1);
+#else
+  struct addrinfo hints{};
+  hints.ai_family = AF_INET;
+  struct addrinfo *r = nullptr;
+  ok = lwip_getaddrinfo("pool.ntp.org", nullptr, &hints, &r) == 0 && r != nullptr;
+  if (ok) {
+    inet_ntoa_r(((struct sockaddr_in *) r->ai_addr)->sin_addr, res, sizeof(res));
+    lwip_freeaddrinfo(r);
+  }
+#endif
+  ESP_LOGI("dnsprobe", "dns0=%s dns1=%s lookup=%s %s", b0, b1, ok ? "ok" : "FAIL", res);
+}

.agents/skills/hardware-test/scripts/manual_ip/plat-esp32.yaml

@@ -0,0 +1,4 @@
+esp32:
+  board: esp32dev
+  framework:
+    type: esp-idf

.agents/skills/hardware-test/scripts/manual_ip/plat-esp8266.yaml

@@ -0,0 +1,2 @@
+esp8266:
+  board: d1_mini

.agents/skills/hardware-test/scripts/manual_ip/run-ap.sh

@@ -0,0 +1,30 @@
+#!/bin/bash
+# usage: run-ap.sh LABEL   the DUT runs the fallback AP on 10.42.7.1, the other board joins it and fetches its page
+# Copy this directory into a scratch directory and run it there; no secrets are needed.
+# env: TREE (esphome checkout for the DUT), CLIENT_TREE (checkout for the client, default TREE), DUT_PORT, DUT_PLAT,
+# CLIENT_PORT, CLIENT_PLAT (esp32 or esp8266), SC (path to serial_capture.py), PY (default $TREE/venv/bin/python)
+D=$(cd "$(dirname "$0")" && pwd); L=${1:?label}; T=${TREE:?set TREE}; CT=${CLIENT_TREE:-$T}
+PY=${PY:-$T/venv/bin/python}; SC=${SC:?set SC}; DP=${DUT_PORT:?set DUT_PORT}; CP=${CLIENT_PORT:?set CLIENT_PORT}
+N=mip-ap-$L; CN=mip-client
+up() { (cd "$1" && PYTHONUNBUFFERED=1 timeout 1500 $PY -m esphome -s name "$2" -s plat "$3" run --no-logs "$D/$4" \
+  --device "$5" > "$D/$2-upload.log" 2>&1) || { echo "FAIL: $2 build or upload"; exit 1; }; }
+up "$T" "$N" "${DUT_PLAT:-esp32}" ap.yaml "$DP"
+up "$CT" "$CN" "${CLIENT_PLAT:-esp8266}" client.yaml "$CP"
+rm -f "$D/$N-serial.txt" "$D/$CN-$N-serial.txt"
+$PY "$SC" "$DP" 120 "$D/$N-serial.txt" --reset > /dev/null 2>&1 & C1=$!
+$PY "$SC" "$CP" 120 "$D/$CN-$N-serial.txt" --reset > /dev/null 2>&1 & C2=$!
+wait $C1; R1=$?; wait $C2; R2=$?
+strip() { LC_ALL=C sed 's/\x1b\[[0-9;]*m//g' | sort -u -k2; }
+echo "--- DUT"; LC_ALL=C grep -a -E "Manual IP|Starting fallback|IP Address|Setting up AP|rst cause|Guru|abort|panic" \
+  "$D/$N-serial.txt" | strip | head -8
+echo "--- client"; LC_ALL=C grep -a -E "IP Address:|Gateway:|Subnet:" "$D/$CN-$N-serial.txt" | strip | head -4
+ok=$(grep -a -c 'status=200' "$D/$CN-$N-serial.txt"); bad=$(grep -a -c 'GET http://10.42.7.1/ failed' "$D/$CN-$N-serial.txt")
+echo "client GET ok=$ok failed=$bad"
+fail=0
+[ "$R1" = 0 ] && [ -s "$D/$N-serial.txt" ] || { echo "FAIL: DUT capture failed or empty (rc $R1)"; fail=1; }
+[ "$R2" = 0 ] && [ -s "$D/$CN-$N-serial.txt" ] || { echo "FAIL: client capture failed or empty (rc $R2)"; fail=1; }
+grep -a -q -E "IP Address: +10\.42\.7\.1\b" "$D/$N-serial.txt" || { echo "FAIL: AP address is not 10.42.7.1"; fail=1; }
+grep -a -q -E "IP Address: +10\.42\.7\." "$D/$CN-$N-serial.txt" || { echo "FAIL: client never joined"; fail=1; }
+[ "$ok" -gt 0 ] || { echo "FAIL: no page fetched"; fail=1; }
+grep -a -q -E "Guru|abort\(\)|panic|[eE]xception \(|Soft WDT reset|wdt reset" "$D/$N-serial.txt" && { echo "FAIL: DUT crash"; fail=1; }
+[ $fail = 0 ] && echo PASS; exit $fail

.agents/skills/hardware-test/scripts/manual_ip/run-sta.sh

@@ -0,0 +1,24 @@
+#!/bin/bash
+# usage: run-sta.sh LABEL   flash sta.yaml, capture 75 s, ping the manual address, check the dump and the DNS probe
+# Copy this directory into a scratch directory with a secrets.yaml (wifi_ssid, wifi_password) and run it there.
+# env: TREE (esphome checkout), PORT (board serial port), PLAT (esp32 or esp8266), IP (the board's own DHCP lease),
+# GATEWAY, SUBNET, DNS1 (the network's values), SC (path to serial_capture.py), PY (default $TREE/venv/bin/python)
+D=$(cd "$(dirname "$0")" && pwd); L=${1:?label}; T=${TREE:?set TREE}; PY=${PY:-$T/venv/bin/python}
+P=${PORT:?set PORT}; IP=${IP:?set IP}; SC=${SC:?set SC}; N=mip-sta-$L; C=$D/$N-serial.txt
+(cd "$T" && PYTHONUNBUFFERED=1 timeout 1500 $PY -m esphome -s name "$N" -s plat "${PLAT:-esp32}" -s ip "$IP" \
+  -s gateway "${GATEWAY:?set GATEWAY}" -s subnet "${SUBNET:?set SUBNET}" -s dns1 "${DNS1:?set DNS1}" \
+  run --no-logs "$D/sta.yaml" --device "$P" > "$D/$N-upload.log" 2>&1) || { echo "FAIL: build or upload"; exit 1; }
+rm -f "$C"
+$PY "$SC" "$P" 75 "$C" --reset > /dev/null 2>&1 & CAP=$!
+# ping exits 0 when at least one reply came back, on macOS and Linux alike.
+sleep 50; PING=$(ping -c 10 -i 0.3 "$IP" 2>&1); PING_RC=$?; wait $CAP; CAPRC=$?
+LC_ALL=C grep -a -E "Manual IP|IP Address:|Gateway:|Subnet:|DNS1:|DNS2:|dnsprobe|rst cause|Guru|abort|panic" "$C" \
+  | LC_ALL=C sed 's/\x1b\[[0-9;]*m//g' | sort -u -k2 | head -14
+echo "ping: $(echo "$PING" | grep -E 'transmitted')"
+fail=0
+[ "$CAPRC" = 0 ] && [ -s "$C" ] || { echo "FAIL: serial capture failed or empty (rc $CAPRC)"; fail=1; }
+grep -a -q -E "dns0=$DNS1 dns1=9\.9\.9\.9 lookup=ok" "$C" || { echo "FAIL: DNS servers or lookup"; fail=1; }
+grep -a -q -E "IP Address: +$IP\b" "$C" || { echo "FAIL: address is not $IP"; fail=1; }
+[ $PING_RC = 0 ] || { echo "FAIL: no ping reply"; fail=1; }
+grep -a -q -E "Guru|abort\(\)|panic|[eE]xception \(|Soft WDT reset|wdt reset" "$C" && { echo "FAIL: crash"; fail=1; }
+[ $fail = 0 ] && echo PASS; exit $fail

.agents/skills/hardware-test/scripts/manual_ip/sta.yaml

@@ -0,0 +1,62 @@
+# Station manual IP. The real network is the second table row and inherits the top level manual_ip; set ip to the
+# board's own DHCP lease and dns2 to a server DHCP does not hand out, so the logs prove the manual values are used.
+# run-sta.sh passes name, plat, ip, gateway, subnet and dns1 with -s; dns2 stays 9.9.9.9, which it checks for.
+substitutions:
+  name: mip-sta
+  plat: esp32
+  ip: 192.168.1.50
+  gateway: 192.168.1.1
+  subnet: 255.255.255.0
+  dns1: 192.168.1.1
+  dns2: 9.9.9.9
+
+esphome:
+  name: ${name}
+  includes:
+    - dnsprobe.hpp
+
+packages:
+  plat: !include plat-${plat}.yaml
+
+logger:
+  level: VERBOSE
+  logs:
+    scheduler: DEBUG
+    component: DEBUG
+    api.service: DEBUG
+    api.connection: DEBUG
+    api.socket: DEBUG
+    socket: DEBUG
+    mdns: DEBUG
+
+wifi:
+  manual_ip:
+    static_ip: ${ip}
+    gateway: ${gateway}
+    subnet: ${subnet}
+    dns1: ${dns1}
+    dns2: ${dns2}
+  networks:
+    - ssid: mip-nowhere
+      password: nowhere123
+      manual_ip:
+        static_ip: 10.9.9.9
+        gateway: 10.9.9.1
+        subnet: 255.255.255.0
+        dns1: 10.9.9.1
+    - ssid: !secret wifi_ssid
+      password: !secret wifi_password
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+interval:
+  - interval: 20s
+    then:
+      - if:
+          condition: wifi.connected
+          then:
+            - lambda: dns_probe();

.agents/skills/hardware-test/scripts/ota_timing/apiinfo.py

@@ -0,0 +1,43 @@
+"""Check over the API that each board runs the expected build.
+Usage: apiinfo.py HOST[=BUILD_INFO_JSON]...
+Prints name and compilation time per host. With a build_info.json (from the build directory), the board's
+compilation time must equal that build's build_time_str. Exits 1 when a host fails or runs another build.
+env: API_KEY (the boards' api encryption key, when they have one)"""
+
+import asyncio
+import json
+import os
+from pathlib import Path
+import sys
+
+from aioesphomeapi import APIClient, APIConnectionError
+
+
+async def info(host: str) -> tuple[str, str]:
+    cli = APIClient(host, 6053, None, noise_psk=os.environ.get("API_KEY"))
+    await cli.connect(login=True)
+    try:
+        i = await cli.device_info()
+    finally:
+        await cli.disconnect()
+    return i.name, i.compilation_time
+
+
+bad = 0
+for arg in sys.argv[1:]:
+    host, _, build_info = arg.partition("=")
+    try:
+        name, built = asyncio.run(asyncio.wait_for(info(host), 30))
+    except (APIConnectionError, TimeoutError, OSError) as err:
+        print("API", host, "ERROR", repr(err), flush=True)
+        bad += 1
+        continue
+    want = (
+        json.loads(Path(build_info).read_text(encoding="utf-8"))["build_time_str"]
+        if build_info
+        else None
+    )
+    ok = want is None or built == want
+    bad += not ok
+    print("API", host, name, built, "" if ok else f"MISMATCH want {want}", flush=True)
+sys.exit(1 if bad else 0)

.agents/skills/hardware-test/scripts/ota_timing/inflate_bench/bench-bk72xx.yaml

@@ -0,0 +1,42 @@
+# Inflate benchmark: run gen_streams.py first, then bench.sh. Results go out over syslog to syslog_ip:5514.
+substitutions:
+  name: inflbench-bk
+  board: cb3s
+
+esphome:
+  name: ${name}
+  includes:
+    - bench_streams.hpp
+    - bench.hpp
+
+bk72xx:
+  board: ${board}
+
+logger:
+  hardware_uart: UART1
+
+wifi:
+  ssid: !secret wifi_ssid
+  password: !secret wifi_password
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+udp:
+  addresses:
+    - !secret syslog_ip
+
+time:
+  - platform: sntp
+
+syslog:
+  port: 5514
+
+interval:
+  - interval: 90s
+    then:
+      - delay: 20s
+      - lambda: run_inflate_bench();

.agents/skills/hardware-test/scripts/ota_timing/inflate_bench/bench-esp32s3.yaml

@@ -0,0 +1,44 @@
+# Inflate benchmark: run gen_streams.py first, then bench.sh. Results go out over syslog to syslog_ip:5514.
+substitutions:
+  name: inflbench-s3
+
+esphome:
+  name: ${name}
+  includes:
+    - bench_streams.hpp
+    - bench.hpp
+
+esp32:
+  board: esp32-s3-devkitc-1
+  variant: esp32s3
+  framework:
+    type: esp-idf
+
+logger:
+  hardware_uart: USB_SERIAL_JTAG
+
+wifi:
+  ssid: !secret wifi_ssid
+  password: !secret wifi_password
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+udp:
+  addresses:
+    - !secret syslog_ip
+
+time:
+  - platform: sntp
+
+syslog:
+  port: 5514
+
+interval:
+  - interval: 90s
+    then:
+      - delay: 20s
+      - lambda: run_inflate_bench();

.agents/skills/hardware-test/scripts/ota_timing/inflate_bench/bench.hpp

@@ -0,0 +1,54 @@
+#pragma once
+// On-device inflate speed for the ESPHome OTA decoder. Targets the decoder interface with dest_limit (dev as of
+// this writing); once the interface without dest_limit lands (#21641), drop the dest_limit line.
+#include "esphome/components/esphome/ota/ota_esphome_inflate.h"
+#include "esphome/core/application.h"
+#include "esphome/core/hal.h"
+#include "esphome/core/log.h"
+#include "bench_streams.hpp"
+
+static uint8_t bench_ring[32768];
+static OtaInflateState bench_state;
+
+// LibreTiny defines micros as a macro; the parentheses call the ESPHome function.
+inline uint32_t bench_now() { return (esphome::micros) (); }
+
+inline void run_inflate_bench() {
+  unsigned expected = 0;
+  for (const auto &c : BENCH_CASES) {
+    for (unsigned ring : {4096u, 8192u, 16384u, 32768u})
+      expected += ring >= c.enc_window;
+  }
+  ESP_LOGI("bench", "BENCHSTART cases=%u", expected);
+  for (const auto &c : BENCH_CASES) {
+    for (unsigned ring : {4096u, 8192u, 16384u, 32768u}) {
+      if (ring < c.enc_window)
+        continue;
+      uint32_t best = UINT32_MAX;
+      size_t out = 0;
+      int res = 0;
+      for (int rep = 0; rep < 3; rep++) {
+        ota_inflate_init(&bench_state, bench_ring, ring);
+        bench_state.source = c.data;
+        bench_state.source_limit = c.data + c.len;
+        bench_state.source_read_cb = [](OtaInflateState *) -> int { return -1; };
+        out = 0;
+        uint32_t t0 = bench_now();
+        do {
+          bench_state.dest = bench_ring;
+          bench_state.dest_limit = bench_ring + ring;
+          res = ota_inflate(&bench_state);
+          out += bench_state.dest - bench_ring;
+          esphome::App.feed_wdt();
+        } while (res == OTA_INFLATE_OK);
+        uint32_t dt = bench_now() - t0;
+        if (dt < best)
+          best = dt;
+      }
+      ESP_LOGI("bench", "%s enc=%u ring=%u in=%u out=%u/%u res=%d best_us=%u out_kBps=%u in_kBps=%u", c.name,
+               c.enc_window, ring, (unsigned) c.len, (unsigned) out, (unsigned) c.raw, res, (unsigned) best,
+               (unsigned) ((uint64_t) out * 1000 / best), (unsigned) ((uint64_t) c.len * 1000 / best));
+    }
+  }
+  ESP_LOGI("bench", "BENCHDONE");
+}

.agents/skills/hardware-test/scripts/ota_timing/inflate_bench/bench.sh

@@ -0,0 +1,42 @@
+#!/bin/bash
+# usage: bench.sh CONFIG DEVICE   build and upload CONFIG (bench-bk72xx or bench-esp32s3) to DEVICE (OTA address or
+# serial port), then collect one benchmark pass over syslog. Run gen_streams.py in this directory first.
+# Copy this directory into a scratch directory with a secrets.yaml (wifi_ssid, wifi_password, syslog_ip = this
+# host's address) and run it there. env: TREE (esphome checkout under test), PY (checkout python), WAIT (seconds to
+# wait for a pass, default 300)
+D=$(cd "$(dirname "$0")" && pwd); C=${1:?config name}; DEV=${2:?device}; T=${TREE:?set TREE}
+PY=${PY:-$T/venv/bin/python}; WAIT=${WAIT:-300}; OUT=$D/$C-syslog.txt
+[ -f "$D/bench_streams.hpp" ] || { echo "FAIL: run gen_streams.py first"; exit 1; }
+(cd "$T" && PYTHONUNBUFFERED=1 timeout 1500 $PY -m esphome run --no-logs "$D/$C.yaml" --device "$DEV" > "$D/$C-upload.log" 2>&1) \
+  || { echo "FAIL: build or upload failed, see $C-upload.log"; exit 1; }
+$PY - "$OUT" "$WAIT" << 'PYEOF'
+import socket
+import sys
+import time
+
+s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
+s.bind(("0.0.0.0", 5514))
+s.settimeout(1)
+end = time.time() + int(sys.argv[2])
+started = False
+with open(sys.argv[1], "w", encoding="utf-8") as out:
+    while time.time() < end:
+        try:
+            d, _ = s.recvfrom(2048)
+        except TimeoutError:
+            continue
+        line = d.decode(errors="replace").strip()
+        # keep one whole pass: lines from a pass already under way are dropped
+        started = started or "BENCHSTART" in line
+        if started and "bench" in line:
+            print(line, file=out, flush=True)
+            if "BENCHDONE" in line:
+                break
+PYEOF
+grep -q BENCHDONE "$OUT" || { echo "FAIL: no complete pass within $WAIT s"; exit 1; }
+grep -o 'bench:[0-9]*\]: .*' "$OUT"
+BAD=$(grep 'best_us=' "$OUT" | grep -v 'res=1 ' | wc -l | tr -d ' ')
+SHORT=$(grep -o 'out=[0-9]*/[0-9]*' "$OUT" | awk -F'[=/]' '$2 != $3' | wc -l | tr -d ' ')
+WANT=$(grep -o 'BENCHSTART cases=[0-9]*' "$OUT" | cut -d= -f2); GOT=$(grep -c 'best_us=' "$OUT")
+[ "$BAD" = 0 ] && [ "$SHORT" = 0 ] && [ "$GOT" = "$WANT" ] && echo "PASS" \
+  || { echo "FAIL: $BAD cases did not finish, $SHORT short outputs, $GOT of $WANT results (lost syslog lines)"; exit 1; }

.agents/skills/hardware-test/scripts/ota_timing/inflate_bench/gen_streams.py

@@ -0,0 +1,43 @@
+"""Write bench_streams.hpp: one 128 KB slice per image, raw deflate at level 9 with 4 KB and 32 KB windows.
+Usage: gen_streams.py NAME=IMAGE[@OFFSET]... [--windows 12,15]
+Each image gives one case per window. Keep the total near 130 KB on a BK7231N: an image with about 330 KB of extra
+incompressible data failed its OTA ("Writing OTA data to flash memory failed"); use one image and --windows 12 there."""
+
+import argparse
+from pathlib import Path
+import zlib
+
+SLICE = 128 * 1024
+ap = argparse.ArgumentParser()
+ap.add_argument("images", nargs="+")
+ap.add_argument("--windows", default="12,15")
+args = ap.parse_args()
+arrays, cases, total = [], [], 0
+for spec in args.images:
+    name, _, rest = spec.partition("=")
+    path, _, offset = rest.partition("@")
+    raw = Path(path).read_bytes()[int(offset or 0) :][:SLICE]
+    for wbits in (int(w) for w in args.windows.split(",")):
+        data = zlib.compress(raw, 9, -wbits)
+        total += len(data)
+        var = f"{name}_w{wbits}"
+        arrays.append(
+            f"static const uint8_t {var}[{len(data)}] = {{{','.join(map(str, data))}}};"
+        )
+        cases.append(f'{{"{name}", {var}, sizeof({var}), {len(raw)}u, {1 << wbits}u}}')
+        print(f"{var}: {len(raw)} -> {len(data)} bytes")
+head = "#pragma once\n#include <cstddef>\n#include <cstdint>\n"
+case_struct = (
+    "struct BenchCase {\n  const char *name;\n  const uint8_t *data;\n  size_t len;\n  size_t raw;\n"
+    "  unsigned enc_window;\n};\n"
+)
+body = (
+    "\n".join(arrays)
+    + "\n"
+    + case_struct
+    + "static const BenchCase BENCH_CASES[] = {"
+    + ", ".join(cases)
+    + "};\n"
+)
+Path("bench_streams.hpp").write_text(head + body, encoding="utf-8")
+print(f"total {total} bytes of streams")

.agents/skills/hardware-test/scripts/ota_timing/otatime.py

@@ -0,0 +1,88 @@
+"""Time one ESPHome OTA upload and print one RESULT json line.
+Usage: otatime.py HOST IMAGE MODE   (MODE: deflate, raw, deflate-tcp or raw-tcp)
+env: OTA_PORT (default 3232), OTA_PASSWORD, API_KEY (a board with an api encryption key also encrypts OTA),
+TAG (label stored in the result)
+Run it with the checkout under test first on PYTHONPATH, so its espota2 is the client being measured.
+raw: the client leaves deflate out of its feature offer. Never patch the deflate branch off after the offer: once
+the device accepted deflate it expects a deflate stream. -tcp: the link is reported clean, so the client asks to
+stay on TCP. Clearing the UDP feature bit instead desyncs the handshake ("receiving UDP token response: timed out")."""
+
+import json
+import logging
+import os
+from pathlib import Path
+import sys
+import time
+
+import esphome.espota2 as espota
+
+host, image, mode = sys.argv[1], Path(sys.argv[2]), sys.argv[3]
+# Assigning a missing attribute would silently do nothing, so a renamed name must stop the run
+for name in ("CLIENT_FEATURE_SUPPORTS_DEFLATE", "_loss_reason"):
+    if not hasattr(espota, name):
+        sys.exit(f"espota2 in this tree has no {name}; update otatime.py")
+if mode.startswith("raw"):
+    espota.CLIENT_FEATURE_SUPPORTS_DEFLATE = 0
+if mode.endswith("-tcp"):
+    espota._loss_reason = lambda *a: None  # noqa: SLF001
+marks: list[tuple[float, float]] = []
+
+
+class Recorder:
+    """Stands in for the progress bar and records when each fraction went out."""
+
+    def __init__(self, *args) -> None:
+        pass
+
+    def update(self, progress: float) -> None:
+        marks.append((time.perf_counter(), progress))
+
+    def done(self) -> None:
+        pass
+
+
+espota.ProgressBar = Recorder
+lines: list[str] = []
+
+
+class Keep(logging.Handler):
+    def emit(self, record: logging.LogRecord) -> None:
+        lines.append(record.getMessage())
+
+
+logging.basicConfig(level=logging.INFO)
+logging.getLogger().addHandler(Keep())
+t0 = time.perf_counter()
+rc, _ = espota.run_ota(
+    host,
+    int(os.environ.get("OTA_PORT", "3232")),
+    os.environ.get("OTA_PASSWORD"),
+    image,
+    noise_psk=os.environ.get("API_KEY"),
+)
+total = time.perf_counter() - t0
+deciles = []
+if marks:
+    start = marks[0][0]
+    for q in range(1, 11):
+        hit = next((t for t, f in marks if f >= q / 10 - 1e-9), None)
+        deciles.append(round(hit - start, 2) if hit is not None else None)
+result = {
+    "board": os.environ.get("TAG", ""),
+    "mode": mode,
+    "rc": rc,
+    "total": round(total, 2),
+    "took": next((x for x in lines if x.startswith("Update took")), ""),
+    "size": next((x for x in lines if "Compressed to" in x), ""),
+    "udp": any("Using UDP" in x for x in lines),
+    "deciles": deciles,
+}
+# The log must agree with the mode, or the row would be counted under the wrong mode
+if mode.startswith("raw") and result["size"]:
+    result["rc"] = rc = rc or 1
+    print("raw mode upload was compressed anyway", file=sys.stderr)
+if mode.endswith("-tcp") and result["udp"]:
+    result["rc"] = rc = rc or 1
+    print("-tcp mode upload switched to UDP", file=sys.stderr)
+print("RESULT", json.dumps(result), flush=True)
+sys.exit(1 if rc else 0)

.agents/skills/hardware-test/scripts/ota_timing/run.sh

@@ -0,0 +1,65 @@
+#!/bin/bash
+# usage: run.sh N PHASE...   OTA upload timing A/B: per phase build every board, one warm-up upload per board
+# (installs the phase's build, not timed), an API check that each board booted that build, then N rounds of every
+# MODE on every board, interleaved so the boards share the same air. summ.py prints medians at the end.
+# Layout: $D/PHASE/NAME.yaml per phase and board (with its secrets.yaml); outputs go to $D/res.txt.
+# env: BOARDS ("NAME,HOST,OTA_PORT ..."; add an ESP32 on the same Wi-Fi as a network control), TREE_<PHASE>
+# (esphome checkout for each phase, e.g. TREE_base, TREE_pr), PY (python with esphome's requirements),
+# MODES (default "deflate-tcp raw-tcp"), SETTLE (seconds after a board answers again, default 15), API_KEY (when the
+# boards have an api encryption key; it also encrypts their OTA, so all boards in one run need the same key)
+D=$(cd "$(dirname "$0")" && pwd); N=${1:?rounds}; shift; [ $# -gt 0 ] || { echo "usage: run.sh N PHASE..."; exit 2; }
+B=${BOARDS:?set BOARDS}; PY=${PY:?set PY}; MODES=${MODES:-deflate-tcp raw-tcp}; SETTLE=${SETTLE:-15}
+wait_up() {  # wait_up HOST PORT: up to 5 minutes for the OTA port to answer
+  local i
+  for i in $(seq 60); do
+    "$PY" -c "import socket,sys; socket.create_connection((sys.argv[1], int(sys.argv[2])), 2).close()" "$1" "$2" 2>/dev/null \
+      && { sleep "$SETTLE"; return 0; }
+    sleep 5
+  done
+  return 1
+}
+up() {  # up PHASE NAME HOST PORT MODE TAG OUTFILE; returns 1 unless the upload succeeded
+  local tree=TREE_$1 out=""
+  if wait_up "$3" "$4"; then
+    out=$(TAG=$1-$2-$6 OTA_PORT=$4 PYTHONPATH=${!tree} timeout 900 "$PY" "$D/otatime.py" "$3" \
+      "$(cat "$D/img-$1-$2.txt")" "$5" 2>> "$D/stderr-$1.log" | grep RESULT)
+  else
+    echo "FAIL: $3 not answering on $4"
+  fi
+  # no RESULT line (board gone, otatime.py crashed or timed out): record a failed upload so summ.py counts it
+  [ -n "$out" ] || out="RESULT {\"board\": \"$1-$2-$6\", \"mode\": \"$5\", \"rc\": -1, \"total\": 0, \"took\": \"\", \"size\": \"\", \"udp\": false, \"deciles\": []}"
+  echo "$out" | tee -a "$7"
+  [[ $out == *'"rc": 0,'* ]]
+}
+# results from an earlier run would fill in for missing rows: move them aside
+for f in res.txt warm.txt api.txt; do [ -f "$D/$f" ] && mv "$D/$f" "$D/$f.$(date +%Y%m%d-%H%M%S)"; done
+for ph in "$@"; do
+  tree=TREE_$ph; [ -n "${!tree}" ] || { echo "FAIL: set $tree"; exit 1; }
+  checks=()
+  for b in $B; do
+    IFS=, read -r n h p <<< "$b"
+    (cd "${!tree}" && PYTHONUNBUFFERED=1 timeout 1500 "$PY" -m esphome compile "$D/$ph/$n.yaml") > "$D/build-$ph-$n.log" 2>&1 \
+      || { echo "FAIL: build $ph/$n"; exit 1; }
+    # the image the CLI uploads (CORE.firmware_bin): ESP-IDF build, LibreTiny UF2, else the PlatformIO bin
+    bd=$D/$ph/.esphome/build/$n; : > "$D/img-$ph-$n.txt"
+    for f in "$bd/build/$n.bin" "$bd/.pioenvs/$n/firmware.uf2" "$bd/.pioenvs/$n/firmware.bin"; do
+      [ -f "$f" ] && { echo "$f" > "$D/img-$ph-$n.txt"; break; }
+    done
+    [ -s "$D/img-$ph-$n.txt" ] || { echo "FAIL: no firmware image under $bd"; exit 1; }
+    case $(cat "$D/img-$ph-$n.txt") in
+      *.uf2) PYTHONPATH=${!tree} "$PY" "$D/trim_uf2.py" "$(cat "$D/img-$ph-$n.txt")" > "$D/img-$ph-$n.tmp" \
+        && mv "$D/img-$ph-$n.tmp" "$D/img-$ph-$n.txt" || { echo "FAIL: trim_uf2.py $ph/$n"; exit 1; } ;;
+    esac
+    checks+=("$h=$D/$ph/.esphome/build/$n/build_info.json")
+  done
+  for b in $B; do IFS=, read -r n h p <<< "$b"; up "$ph" "$n" "$h" "$p" deflate-tcp warm "$D/warm.txt" || exit 1; done
+  for b in $B; do IFS=, read -r n h p <<< "$b"; wait_up "$h" "$p"; done
+  "$PY" "$D/apiinfo.py" "${checks[@]}" | tee -a "$D/api.txt"; [ "${PIPESTATUS[0]}" = 0 ] || { echo "FAIL: $ph did not boot everywhere"; exit 1; }
+  for i in $(seq "$N"); do
+    for m in $MODES; do
+      for b in $B; do IFS=, read -r n h p <<< "$b"; up "$ph" "$n" "$h" "$p" "$m" "r$i" "$D/res.txt"; done
+    done
+  done
+  echo "PHASE DONE $ph"
+done
+"$PY" "$D/summ.py" "$D/res.txt" "$N"

.agents/skills/hardware-test/scripts/ota_timing/summ.py

@@ -0,0 +1,43 @@
+"""Median upload time per board and mode from otatime.py RESULT lines.
+Usage: summ.py RESULTS_FILE [ROUNDS]
+Exits 1 when any upload failed, an upload time could not be read, or (with ROUNDS) a board and mode has fewer rows."""
+
+import json
+from pathlib import Path
+import re
+import statistics
+import sys
+
+rows = [
+    json.loads(line.split("RESULT ", 1)[1])
+    for line in Path(sys.argv[1]).read_text(encoding="utf-8").splitlines()
+    if "RESULT " in line
+]
+by: dict[tuple[str, str], list[dict]] = {}
+for r in rows:
+    m = re.search(r"upload ([\d.]+)", r["took"])
+    r["up"] = float(m[1]) if m else None
+    # TAG is PHASE-BOARD-ROUND; group per phase and board
+    by.setdefault((r["board"].rsplit("-", 1)[0], r["mode"]), []).append(r)
+rounds = int(sys.argv[2]) if len(sys.argv) > 2 else 0
+failed = unparsed = short = 0
+for (board, mode), v in sorted(by.items()):
+    ok = sorted(r["up"] for r in v if r["rc"] == 0 and r["up"] is not None)
+    fails = sum(r["rc"] != 0 for r in v)
+    failed += fails
+    unparsed += sum(r["rc"] == 0 and r["up"] is None for r in v)
+    short += len(v) < rounds
+    udp = sum(r["udp"] for r in v if r["rc"] == 0)
+    med = statistics.median(ok) if ok else None
+    print(
+        f"{board:24} {mode:12} n={len(v)} fail={fails} udp={udp} median={med} upload s: {ok}"
+    )
+if not rows:
+    print("FAIL: no results")
+elif failed or unparsed or short:
+    print(
+        f"FAIL: {failed} failed uploads, {unparsed} without an upload time, {short} groups under {rounds} rows"
+    )
+else:
+    print("PASS")
+sys.exit(0 if rows and not (failed or unparsed or short) else 1)

.agents/skills/hardware-test/scripts/ota_timing/trim_uf2.py

@@ -0,0 +1,24 @@
+"""Write the trimmed LibreTiny UF2 the CLI would send (esphome#21642) next to firmware.uf2 and print its path.
+Usage: trim_uf2.py FIRMWARE_UF2   (run with the phase's checkout on PYTHONPATH)
+Prints the input path unchanged when the tree has no device_ota_uf2() or there is nothing to drop."""
+
+from pathlib import Path
+import sys
+
+src = Path(sys.argv[1])
+try:
+    from esphome.components.libretiny.uf2 import device_ota_uf2
+except ModuleNotFoundError as err:
+    if err.name != "esphome.components.libretiny.uf2":
+        raise
+    print("no device_ota_uf2() in this tree, sending the full UF2", file=sys.stderr)
+    print(src)
+    sys.exit(0)
+data = src.read_bytes()
+if (trimmed := device_ota_uf2(data)) is None:
+    print(src)
+    sys.exit(0)
+out = src.with_name("firmware.ota.uf2")
+out.write_bytes(trimmed)
+print(f"trimmed {len(data)} -> {len(trimmed)} bytes", file=sys.stderr)
+print(out)

.agents/skills/hardware-test/scripts/prometheus/datetime_metrics.py

@@ -0,0 +1,71 @@
+"""Print the date, time and datetime rows of /metrics, set new values over the API, print them again.
+
+usage: datetime_metrics.py HOST   env: API_KEY
+Sets every date to 2040-03-04, every time to 23:59:59 and every datetime to 2040-03-04 05:06:07 UTC (past 2038 on
+purpose). Values are printed with their UTC reading: dates are local midnight, times are ms since midnight and
+datetimes are ms since the epoch.
+"""
+
+import asyncio
+from datetime import UTC, datetime
+import os
+import re
+import sys
+import urllib.request
+
+from aioesphomeapi import APIClient, DateInfo, DateTimeInfo, TimeInfo
+
+HOST = sys.argv[1]
+
+
+def metrics(tag: str) -> int:
+    with urllib.request.urlopen(f"http://{HOST}/metrics", timeout=10) as resp:  # noqa: S310
+        body = resp.read().decode()
+    rows = [
+        line
+        for line in body.splitlines()
+        if re.match(r"esphome_(date|time|datetime)_", line)
+    ]
+    print(f"--- {tag}: {len(rows)} rows")
+    parsed = 0
+    for line in rows:
+        m = re.match(r'(esphome_\w+)\{.*name="([^"]+)"\} (-?\d+)$', line)
+        if not m:
+            print(f"  UNPARSED {line}")
+            continue
+        parsed += 1
+        metric, name, value = m.group(1), m.group(2), int(m.group(3))
+        utc = ""
+        if metric in ("esphome_date_value", "esphome_datetime_value"):
+            utc = " = " + datetime.fromtimestamp(value / 1000, UTC).isoformat()
+        print(f"  {metric} {name} {value}{utc}")
+    return parsed
+
+
+async def main() -> None:
+    if not metrics("initial"):
+        sys.exit(
+            "NO-DATA: no date, time or datetime rows (a build without them, or no such entities)"
+        )
+    c = APIClient(HOST, 6053, None, noise_psk=os.environ.get("API_KEY"))
+    await c.connect(login=True)
+    ents, _ = await c.list_entities_services()
+    sent = 0
+    for e in ents:
+        sent += isinstance(e, (DateInfo, TimeInfo, DateTimeInfo))
+        if isinstance(e, DateInfo):
+            c.date_command(e.key, 2040, 3, 4)
+        elif isinstance(e, TimeInfo):
+            c.time_command(e.key, 23, 59, 59)
+        elif isinstance(e, DateTimeInfo):
+            c.datetime_command(
+                e.key, int(datetime(2040, 3, 4, 5, 6, 7, tzinfo=UTC).timestamp())
+            )
+    await asyncio.sleep(2)
+    await c.disconnect()
+    if not sent:
+        sys.exit("NO-DATA: the device lists no date, time or datetime entities")
+    metrics("after setting 2040-03-04, 23:59:59 and 2040-03-04 05:06:07 UTC")
+
+
+asyncio.run(main())

.agents/skills/hardware-test/scripts/prometheus/prom-datetime-tz.yaml

@@ -0,0 +1,10 @@
+# Adds a device time zone (UTC-10, no DST) to prom-datetime.yaml; the zone applies without a time sync.
+substitutions:
+  name: prom-datetime-tz
+
+packages:
+  base: !include prom-datetime.yaml
+
+time:
+  - platform: sntp
+    timezone: HST10

.agents/skills/hardware-test/scripts/prometheus/prom-datetime.yaml

@@ -0,0 +1,46 @@
+# D1 mini with three optimistic template datetime entities. -s name NAME; build prom-datetime-tz.yaml for a time zone.
+substitutions:
+  name: prom-datetime
+
+esphome:
+  name: ${name}
+
+esp8266:
+  board: d1_mini
+
+logger:
+
+wifi:
+  ssid: !secret wifi_ssid
+  password: !secret wifi_password
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+web_server:
+  port: 80
+
+datetime:
+  - platform: template
+    name: Date
+    id: test_date
+    type: date
+    optimistic: true
+    initial_value: "2000-1-2"
+  - platform: template
+    name: Time
+    id: test_time
+    type: time
+    optimistic: true
+    initial_value: "12:34:56am"
+  - platform: template
+    name: DateTime
+    id: test_datetime
+    type: datetime
+    optimistic: true
+    initial_value: "2000-1-2 12:34:56"
+
+prometheus:

.agents/skills/hardware-test/scripts/web_handlers/cp.yaml

@@ -0,0 +1,6 @@
+captive_portal:
+button:
+  - platform: template
+    name: Start portal
+    on_press:
+      - lambda: captive_portal::global_captive_portal->start();

.agents/skills/hardware-test/scripts/web_handlers/plat-esp32.yaml

@@ -0,0 +1,4 @@
+esp32:
+  board: esp32dev
+  framework:
+    type: esp-idf

.agents/skills/hardware-test/scripts/web_handlers/plat-esp8266.yaml

@@ -0,0 +1,2 @@
+esp8266:
+  board: d1_mini

.agents/skills/hardware-test/scripts/web_handlers/probehandler.hpp

@@ -0,0 +1,18 @@
+#pragma once
+// A handler registered after setup(), the way an external component would; answers GET /probe.
+#include "esphome/components/web_server_base/web_server_base.h"
+class ProbeHandler : public AsyncWebHandler {
+ public:
+  bool canHandle(AsyncWebServerRequest *request) const override {
+    if (request->method() != HTTP_GET)
+      return false;
+#ifdef USE_ESP32
+    char url_buf[AsyncWebServerRequest::URL_BUF_SIZE];
+    return request->url_to(url_buf) == "/probe";
+#else
+    return request->url() == ESPHOME_F("/probe");
+#endif
+  }
+  void handleRequest(AsyncWebServerRequest *request) override { request->send(200, "text/plain", "probe-ok"); }
+};
+inline ProbeHandler probe_handler;

.agents/skills/hardware-test/scripts/web_handlers/prom.yaml

@@ -0,0 +1,2 @@
+captive_portal:
+prometheus:

.agents/skills/hardware-test/scripts/web_handlers/run.sh

@@ -0,0 +1,65 @@
+#!/bin/bash
+# usage: run.sh LABEL VARIANT   flash webbase.yaml, fetch /probe and /, then a web OTA of the same image
+# Copy this directory into a scratch directory with a secrets.yaml (wifi_ssid, wifi_password) and run it there.
+# VARIANT is ws, cp or prom (see webbase.yaml). The cp variant presses "Start portal" over the API first.
+# env: TREE (esphome checkout), PORT (board serial port), PLAT (esp32 or esp8266), IP (the board's address),
+# SC (path to serial_capture.py), PY (default $TREE/venv/bin/python), SKILL_SCRIPTS (the skill's scripts directory)
+D=$(cd "$(dirname "$0")" && pwd); L=${1:?label}; V=${2:?variant}; T=${TREE:?set TREE}; PY=${PY:-$T/venv/bin/python}
+P=${PORT:?set PORT}; IP=${IP:?set IP}; SC=${SC:?set SC}; PL=${PLAT:-esp32}; N=web-$V-$PL-$L; C=$D/$N-serial.txt
+# api_button.py and ota_timing/apiinfo.py live in the skill's scripts directory; set SKILL_SCRIPTS when this
+# directory was copied elsewhere.
+S=${SKILL_SCRIPTS:-$(cd "$D/.." && pwd)}
+[ -f "$S/api_button.py" ] && [ -f "$S/ota_timing/apiinfo.py" ] || { echo "FAIL: set SKILL_SCRIPTS to the skill's scripts directory"; exit 1; }
+if [ "$PL" = esp32 ]; then BIN=$D/.esphome/build/$N/build/$N.bin; else BIN=$D/.esphome/build/$N/.pioenvs/$N/firmware.bin; fi
+(cd "$T" && PYTHONUNBUFFERED=1 timeout 1500 $PY -m esphome -s name "$N" -s plat "$PL" -s variant "$V" \
+  run --no-logs "$D/webbase.yaml" --device "$P" > "$D/$N-upload.log" 2>&1) || { echo "FAIL: build or upload"; exit 1; }
+rm -f "$C"
+# Longer than the worst case below: 80 s to come up, 120 s of OTA, 80 s to come back.
+$PY "$SC" "$P" 300 "$C" --reset > /dev/null 2>&1 & CAP=$!
+# Wait on the API port: in the cp variant the HTTP server only starts when the portal does, also after the OTA reboot.
+up() {
+  for _ in $(seq 1 40); do
+    $PY -c 'import socket, sys; socket.create_connection((sys.argv[1], 6053), 1).close()' "$IP" 2> /dev/null && return 0
+    sleep 2
+  done
+  return 1
+}
+up || { echo "FAIL: board not reachable"; kill $CAP; exit 1; }; sleep 8
+fail=0
+if [ "$V" = cp ]; then
+  $PY "$S/api_button.py" "$IP" - press "Start portal" || { echo "FAIL: Start portal press"; fail=1; }
+  sleep 2
+  # The captive portal only answers /config.json while it is active.
+  CFG=$(curl -s -m 10 -o /dev/null -w '%{http_code}' "http://$IP/config.json")
+  echo "GET /config.json: $CFG"; [ "$CFG" = 200 ] || { echo "FAIL: captive portal not active"; fail=1; }
+fi
+PROBE=$(curl -s -m 10 -w ' %{http_code}' "http://$IP/probe")
+echo "GET /probe: $PROBE"
+ROOT=$(curl -s -m 10 -o /dev/null -w '%{http_code}' "http://$IP/")
+echo "GET /: $ROOT"
+# prometheus has no page at / (no answer on ESP32, 404 on ESP8266, on dev too)
+[ "$V" = prom ] || [ "$ROOT" = 200 ] || { echo "FAIL: GET /"; fail=1; }
+if [ "$V" = prom ]; then
+  MET=$(curl -s -m 10 -o /dev/null -w '%{http_code}' "http://$IP/metrics")
+  echo "GET /metrics: $MET"; [ "$MET" = 200 ] || { echo "FAIL: /metrics"; fail=1; }
+fi
+# A failed update also answers 200, with "Update Failed!" in the body.
+OTA=$(curl -s -m 120 -F "update=@$BIN" -w ' %{http_code}' "http://$IP/update" | tr -d '\r\n')
+echo "POST /update ($(wc -c < "$BIN" | tr -d ' ') B): $OTA"
+sleep 5
+if up; then
+  sleep 6
+  $PY "$S/ota_timing/apiinfo.py" "$IP=$D/.esphome/build/$N/build_info.json" 2>&1 | tail -1; API=${PIPESTATUS[0]}
+else
+  echo "FAIL: board not back after the web OTA"; fail=1; API=1
+fi
+wait $CAP; CAPRC=$?
+[ "$CAPRC" = 0 ] && [ -s "$C" ] || { echo "FAIL: serial capture failed or empty (rc $CAPRC)"; fail=1; }
+[ "$PROBE" = "probe-ok 200" ] || { echo "FAIL: /probe"; fail=1; }
+case "$OTA" in *"Update Successful!"*" 200") ;; *) echo "FAIL: web OTA"; fail=1;; esac
+[ "$API" = 0 ] || { echo "FAIL: board not on the uploaded build after the web OTA"; fail=1; }
+# The image is the same build, so prove the reboot from the log: setup finishes again after the OTA reboot line.
+grep -a -q "Performing OTA reboot now" "$C" || { echo "FAIL: no OTA reboot in the serial log"; fail=1; }
+sed -n '/Performing OTA reboot now/,$p' "$C" | grep -a -q 'setup() finished' || { echo "FAIL: no boot after the OTA"; fail=1; }
+grep -a -q -E "Guru|abort\(\)|panic|[eE]xception \(|Soft WDT reset|wdt reset" "$C" && { echo "FAIL: crash"; fail=1; }
+[ $fail = 0 ] && echo PASS; exit $fail

.agents/skills/hardware-test/scripts/web_handlers/webbase.yaml

@@ -0,0 +1,36 @@
+# variant: ws (web_server), cp (captive_portal only, started from a button) or prom (captive_portal and prometheus).
+# A GET /probe handler is added after setup, the way an external component would.
+substitutions:
+  name: web-handlers
+  plat: esp32
+  variant: ws
+
+esphome:
+  name: ${name}
+  includes:
+    - probehandler.hpp
+  on_boot:
+    priority: -100
+    then:
+      - lambda: |-
+          web_server_base::global_web_server_base->add_handler(&probe_handler);
+          ESP_LOGI("probe", "probe handler added after setup");
+
+packages:
+  plat: !include plat-${plat}.yaml
+  variant: !include ${variant}.yaml
+
+logger:
+
+wifi:
+  ssid: !secret wifi_ssid
+  password: !secret wifi_password
+  ap:
+    password: webtest123
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+  - platform: web_server

.agents/skills/hardware-test/scripts/web_handlers/ws.yaml

@@ -0,0 +1,2 @@
+web_server:
+  port: 80

.agents/skills/hardware-test/scripts/wifi_restart/an.py

@@ -0,0 +1,95 @@
+"""Summarise the AP restart run: alive lines over syslog per DUT, reboots, gaps, recovery after each AP start.
+Usage: an.py DIR BUILD [RECOVER_S]   reads BUILD-marks.txt, BUILD-syslog.txt, BUILD-bk.txt, BUILD-cb.txt (optional)
+and BUILD-up.txt (the bridge's API logs)
+Exits 1 when a DUT rebooted, crashed, sent nothing, or had no syslog line within RECOVER_S (default 120) of an
+AP start, when the BK7238 capture (or a CB3S capture that was started) is empty, or when the bridge did not switch
+its AP to WPA2/WPA3 mixed after each AP start."""
+
+from itertools import pairwise
+from pathlib import Path
+import re
+import sys
+
+D, b = Path(sys.argv[1]), sys.argv[2]
+RECOVER = int(sys.argv[3]) if len(sys.argv) > 3 else 120
+
+
+def read(suffix: str) -> str | None:
+    f = D / f"{b}-{suffix}.txt"
+    return f.read_text(encoding="utf-8", errors="replace") if f.exists() else None
+
+
+def secs(t: str) -> int:
+    h, m, s = t.split(":")
+    return int(h) * 3600 + int(m) * 60 + int(float(s))
+
+
+marks = [
+    (secs(line.split()[0]), line.split()[3])
+    for line in read("marks").splitlines()
+    if line.strip()
+]
+sl = (read("syslog") or "").splitlines()
+bad = []
+for dev in ("rebk", "recb"):
+    lines = [line for line in sl if re.search(rf"\b{dev}\b", line)]
+    ups = [
+        (secs(line.split()[0]), int(m[1]))
+        for line in lines
+        if (m := re.search(r"up (\d+) s", line))
+    ]
+    if not ups:
+        print(f"== {dev}: no alive lines over syslog")
+        bad.append(f"{dev} silent")
+        continue
+    reboots = sum(1 for (_, u0), (_, u1) in pairwise(ups) if u1 < u0)
+    gaps = [t1 - t0 for (t0, _), (t1, _) in pairwise(ups) if t1 - t0 > 25]
+    print(
+        f"== {dev}: alive lines {len(ups)}, reboots {reboots}, syslog gaps over 25 s: {gaps}"
+    )
+    if reboots:
+        bad.append(f"{dev} rebooted")
+    for t, st in marks:
+        if st != "on":
+            continue
+        after = [x for x in ups if x[0] > t]
+        first = after[0][0] - t if after else None
+        print(
+            f"   ap back at t={t % 86400 // 3600:02d}:{t % 3600 // 60:02d}:{t % 60:02d}, first syslog line {first} s later"
+        )
+        if first is None or first > RECOVER:
+            bad.append(f"{dev} not back {RECOVER} s after an AP start")
+for dev, suffix in (("rebk", "bk"), ("recb", "cb")):
+    if (s := read(suffix)) is None and dev == "recb":
+        continue  # the CB3S capture is optional
+    if not s:
+        bad.append(f"{dev} serial capture missing or empty")
+        continue
+    crashes = len(re.findall(r"(?i)hard fault|exception|abort|watchdog", s))
+    connected = len(re.findall(r"\[wifi:\d+\]: Connected", s))
+    print(
+        f"-- {dev} serial: lines {s.count(chr(10))}, overflow {s.count('queue overflow')}, fallback AP starts "
+        f"{s.count('Starting fallback AP')}, connected {connected}, boots {s.count('ESPHome version')}, crashes {crashes}"
+    )
+    if crashes:
+        bad.append(f"{dev} crashed")
+up = read("up") or ""
+switches = [
+    secs(m[1]) for m in re.finditer(r"\[(\d+:\d+:\d+)[.\d]*\].*PMF capable: 0", up)
+]
+# each AP start needs its own switch between that mark (less a few seconds of clock skew) and the next mark
+ends = [t for t, _ in marks[1:]] + [float("inf")]
+missed = [
+    t
+    for (t, st), end in zip(marks, ends, strict=True)
+    if st == "on" and not any(t - 5 <= s < end for s in switches)
+]
+starts = sum(1 for _, st in marks if st == "on")
+print(
+    f"-- bridge: WPA2/WPA3 switches {len(switches)}, AP starts {starts} ({len(missed)} without a switch), "
+    f"failures {up.count('AP CONFIG FAILED')}"
+)
+if "AP CONFIG FAILED" in up or missed:
+    bad.append("the bridge did not switch its AP to WPA2/WPA3 mixed after every start")
+print("PASS" if not bad else "FAIL: " + ", ".join(bad))
+sys.exit(1 if bad else 0)

.agents/skills/hardware-test/scripts/wifi_restart/re-common.yaml

@@ -0,0 +1,58 @@
+esphome:
+  name: ${name}
+
+logger:
+  level: DEBUG
+
+wifi:
+  networks:
+    - ssid: !secret wisp_up_ssid
+      password: !secret wisp_up_password
+      manual_ip:
+        static_ip: ${ip}
+        gateway: 10.40.0.1
+        subnet: 255.255.255.0
+        dns1: 10.40.0.1
+  power_save_mode: LIGHT
+  ap:
+    ssid: ${name}-fallback
+    password: !secret fallback_ap_password
+
+captive_portal:
+
+web_server:
+  port: 80
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+udp:
+  addresses:
+    - !secret syslog_ip
+time:
+  - platform: sntp
+syslog:
+  port: 5514
+
+interval:
+  - interval: 10s
+    then:
+      - lambda: ESP_LOGI("alive", "up %u s", (unsigned) (millis() / 1000));
+  - interval: 30s
+    then:
+      - if:
+          condition:
+            lambda: |-
+              static bool done = false;
+              if (done || millis() < 19 * 60 * 1000) return false;
+              done = true;
+              return true;
+          then:
+            - lambda: ESP_LOGI("alive", "GOING HOME");
+            - wifi.configure:
+                ssid: !secret wifi_ssid
+                password: !secret wifi_password
+                save: false

.agents/skills/hardware-test/scripts/wifi_restart/rebk.yaml

@@ -0,0 +1,13 @@
+# BK7238 DUT (NiceMCU XH-WB3S), flashed over serial by run.sh
+substitutions:
+  name: rebk
+  ip: 10.40.0.21
+
+packages:
+  common: !include re-common.yaml
+
+bk72xx:
+  board: xh-wb3s
+
+logger:
+  hardware_uart: UART1

.agents/skills/hardware-test/scripts/wifi_restart/recb.yaml

@@ -0,0 +1,13 @@
+# CB3S (BK7231N) DUT, flashed over the air by run.sh; its log port is UART1 (TX1)
+substitutions:
+  name: recb
+  ip: 10.40.0.22
+
+packages:
+  common: !include re-common.yaml
+
+bk72xx:
+  board: cb3s
+
+logger:
+  hardware_uart: UART1

.agents/skills/hardware-test/scripts/wifi_restart/run.sh

@@ -0,0 +1,57 @@
+#!/bin/bash
+# usage: run.sh BUILD   OTA the CB3S, serial flash the BK7238, then switch the upstream AP off and on three times
+# while both DUTs send alive lines over syslog. BUILD is a label (pr, base) for the output files.
+# Copy this directory into a scratch directory with a secrets.yaml (wifi_ssid, wifi_password, wisp_up_ssid,
+# wisp_up_password, fallback_ap_password, syslog_ip, api_key) and run it there. wisp-upstream-wpa3.yaml runs on the
+# ethernet ESP32-S3 upstream; syslog_ip is this host's address as the DUTs reach it through the bridge's NAT.
+# env: TREE (esphome checkout for the DUTs), BK_PORT (BK7238 serial), CB_PORT (CB3S log only serial, optional),
+# CB_HOST (CB3S address to OTA), UPSTREAM (bridge address), SC (path to serial_capture.py), FLASH_BK (path to
+# flash_bk72xx.py), LT (LibreTiny penv python), PY (checkout python), UP_TREE (checkout for the bridge's API logs)
+D=$(cd "$(dirname "$0")" && pwd); b=${1:?build label}; T=${TREE:?set TREE}; PY=${PY:-$T/venv/bin/python}
+BK=${BK_PORT:?set BK_PORT}; CBH=${CB_HOST:?set CB_HOST}; UP=${UPSTREAM:?set UPSTREAM}; SC=${SC:?set SC}
+FB=${FLASH_BK:?set FLASH_BK}; LT=${LT:-$HOME/.platformio/penv/.libretiny/bin/python}; CB=${CB_PORT:-}
+ap() { curl -s -m 5 -X POST -d "" -o /dev/null -w "%{http_code}" "http://$UP/button/ap%20$1/press"; }
+# retry N PAUSE CMD...: run CMD up to N times, PAUSE seconds apart; fails when every attempt failed
+retry() { local n=$1 pause=$2 i; shift 2; for i in $(seq "$n"); do "$@" && return 0; [ "$i" -lt "$n" ] && sleep "$pause"; done; return 1; }
+build() { (cd "$T" && PYTHONUNBUFFERED=1 timeout 1500 $PY -m esphome "$@"); }
+# only this run's ports: a global pkill stops other runs' captures
+pkill -f "[s]erial_capture.py $BK"; [ -n "$CB" ] && pkill -f "[s]erial_capture.py $CB"
+[ "$(ap on)" = 200 ] || { echo "FAIL: upstream bridge did not take ap on"; exit 1; }
+cb_upload() { build run --no-logs "$D/recb.yaml" --device "$CBH" > "$D/cb-$b-upload.log" 2>&1; }
+retry 3 15 cb_upload || { echo "FAIL: CB3S build or OTA after 3 tries"; exit 1; }
+build compile "$D/rebk.yaml" > "$D/bk-$b-build.log" 2>&1 || { echo "FAIL: BK7238 build"; exit 1; }
+UF2=$D/.esphome/build/rebk/.pioenvs/rebk/firmware.uf2
+bk_flash() { $LT "$FB" flash write -d "$BK" "$UF2" > "$D/bk-$b-flash.log" 2>&1; }
+retry 4 5 bk_flash || { echo "FAIL: BK7238 flash after 4 tries"; exit 1; }
+PIDS=()
+$PY - > "$D/$b-syslog.txt" 2>&1 << 'PYEOF' &
+import socket
+import time
+
+s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
+s.bind(("0.0.0.0", 5514))
+s.settimeout(1)
+end = time.time() + 1250
+while time.time() < end:
+    try:
+        d, a = s.recvfrom(2048)
+        print(time.strftime("%H:%M:%S"), a[0], d.decode(errors="replace").strip(), flush=True)
+    except TimeoutError:
+        pass
+PYEOF
+PIDS+=($!)
+$PY "$SC" "$BK" 1240 "$D/$b-bk.txt" --released > /dev/null 2>&1 & PIDS+=($!)
+# an empty file marks the CB3S capture as requested, so an.py fails it if the capture never writes anything
+[ -n "$CB" ] && { : > "$D/$b-cb.txt"; $PY "$SC" "$CB" 1240 "$D/$b-cb.txt" --released > /dev/null 2>&1 & PIDS+=($!); }
+(cd "${UP_TREE:-$T}" && PYTHONUNBUFFERED=1 exec timeout 1240 $PY -m esphome logs "$D/wisp-upstream-wpa3.yaml" --device "$UP" > "$D/$b-up.txt" 2>&1) & PIDS+=($!)
+trap 'kill "${PIDS[@]}" 2>/dev/null' EXIT
+t0=$(date +%s); : > "$D/$b-marks.txt"
+for step in 180:off 330:on 480:off 630:on 780:off 930:on; do
+  t=${step%%:*}; st=${step#*:}
+  while [ $(($(date +%s) - t0)) -lt "$t" ]; do sleep 1; done
+  code=$(ap "$st")
+  echo "$(date +%H:%M:%S) t=$t ap $st $code" | tee -a "$D/$b-marks.txt"
+  [ "$code" = 200 ] || { echo "FAIL: upstream bridge returned $code for ap $st"; exit 1; }
+done
+wait "${PIDS[@]}"; trap - EXIT
+$PY "$D/an.py" "$D" "$b"

.agents/skills/hardware-test/scripts/wifi_restart/wdap_inc.h

@@ -0,0 +1,3 @@
+#pragma once
+#include <esp_wifi.h>
+#include <esp_event.h>

.agents/skills/hardware-test/scripts/wifi_restart/wisp-upstream-wpa3.yaml

@@ -0,0 +1,95 @@
+esphome:
+  name: wisp-upstream
+  includes:
+    - wdap_inc.h
+  on_boot:
+    priority: -100
+    then:
+      - delay: 3s
+      - script.execute: wpa3_ap
+
+esp32:
+  board: esp32-s3-devkitc-1
+  variant: esp32s3
+  flash_size: 16MB
+  framework:
+    type: esp-idf
+
+logger:
+  hardware_uart: USB_SERIAL_JTAG
+
+ethernet:
+  type: W5500
+  clk_pin: GPIO13
+  mosi_pin: GPIO11
+  miso_pin: GPIO12
+  cs_pin: GPIO14
+  interrupt_pin: GPIO10
+  reset_pin: GPIO9
+
+wifi:
+  ap:
+    ssid: !secret wisp_up_ssid
+    password: !secret wisp_up_password
+    ap_mode: WISP
+    manual_ip:
+      static_ip: 10.40.0.1
+      gateway: 10.40.0.1
+      subnet: 255.255.255.0
+
+api:
+  encryption:
+    key: !secret api_key
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+web_server:
+  port: 80
+
+button:
+  - platform: template
+    name: ap off
+    on_press:
+      - wifi.disable:
+  - platform: template
+    name: ap on
+    on_press:
+      - wifi.enable:
+      - delay: 3s
+      - script.execute: wpa3_ap
+
+script:
+  - id: wpa3_ap
+    then:
+      - lambda: |-
+          static bool registered = false;
+          wifi_config_t c{};
+          esp_err_t err = esp_wifi_get_config(WIFI_IF_AP, &c);
+          if (err == ESP_OK) {
+            c.ap.authmode = WIFI_AUTH_WPA2_WPA3_PSK;
+            c.ap.pmf_cfg.required = false;
+            c.ap.sae_pwe_h2e = WPA3_SAE_PWE_BOTH;
+            err = esp_wifi_set_config(WIFI_IF_AP, &c);
+          }
+          // read back: only a running AP that reports WPA2/WPA3 mixed counts as switched
+          wifi_config_t check{};
+          if (err == ESP_OK)
+            err = esp_wifi_get_config(WIFI_IF_AP, &check);
+          if (err != ESP_OK || check.ap.authmode != WIFI_AUTH_WPA2_WPA3_PSK) {
+            ESP_LOGE("wdap", "AP CONFIG FAILED: %d authmode %d", err, (int) check.ap.authmode);
+          } else {
+            ESP_LOGI("wdap", "AP set to WPA2/WPA3 mixed, PMF capable: %d", err);
+          }
+          if (registered) return;
+          registered = true;
+          esp_event_handler_register(WIFI_EVENT, ESP_EVENT_ANY_ID, [](void *, esp_event_base_t, int32_t id, void *data) {
+            if (id == WIFI_EVENT_AP_STACONNECTED) {
+              auto *e = (wifi_event_ap_staconnected_t *) data;
+              ESP_LOGI("wdap", "STA JOIN %02X:%02X:%02X:%02X:%02X:%02X", e->mac[0], e->mac[1], e->mac[2], e->mac[3], e->mac[4], e->mac[5]);
+            } else if (id == WIFI_EVENT_AP_STADISCONNECTED) {
+              auto *e = (wifi_event_ap_stadisconnected_t *) data;
+              ESP_LOGI("wdap", "STA LEAVE %02X:%02X:%02X:%02X:%02X:%02X reason %u", e->mac[0], e->mac[1], e->mac[2], e->mac[3], e->mac[4], e->mac[5], e->reason);
+            }
+          }, nullptr);

.agents/skills/hardware-test/scripts/wisp_ap/dut-always-on.yaml

@@ -0,0 +1,16 @@
+substitutions:
+  name: wispdut-aoi
+  framework: esp-idf
+packages:
+  common: !include dut-common.yaml
+wifi:
+  networks:
+    - ssid: !secret wisp_up_ssid
+      password: !secret wisp_up_password
+  post_connect_roaming: false
+  ap:
+    ssid: !secret dut_ap_ssid
+    password: !secret dut_ap_password
+    ap_mode: ALWAYS_ON
+web_server:
+  port: 80

.agents/skills/hardware-test/scripts/wisp_ap/dut-common.yaml

@@ -0,0 +1,16 @@
+esphome:
+  name: ${name}
+
+esp32:
+  board: esp32dev
+  framework:
+    type: ${framework}
+
+logger:
+  level: DEBUG
+
+api:
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome

.agents/skills/hardware-test/scripts/wisp_ap/dut-wisp.yaml

@@ -0,0 +1,14 @@
+substitutions:
+  name: wispdut-wisp
+  framework: esp-idf
+packages:
+  common: !include dut-common.yaml
+wifi:
+  networks:
+    - ssid: !secret wisp_up_ssid
+      password: !secret wisp_up_password
+  ap:
+    ssid: !secret dut_ap_ssid
+    password: !secret dut_ap_password
+    ap_mode: WISP
+captive_portal:

.agents/skills/hardware-test/scripts/wisp_ap/run.sh

@@ -0,0 +1,36 @@
+#!/bin/bash
+# usage: run.sh DUT START SECONDS T:STATE...   e.g. run.sh dut-wisp on 600 120:off 300:on
+# Builds and flashes DUT.yaml (dut-wisp or dut-always-on) over serial, sets the upstream AP to START, captures the
+# DUT and the client, and switches the upstream AP at each T seconds; sum.py summarises the client per phase.
+# Copy this directory into a scratch directory with a secrets.yaml (wisp_up_ssid, wisp_up_password, dut_ap_ssid,
+# dut_ap_password, lan_probe_url, api_key) and run it there. wisp-upstream.yaml runs on an ethernet ESP32-S3 and
+# wispcli.yaml on a D1 mini; flash both once (CLI_FLASH=1 builds and flashes the client too).
+# env: TREE (esphome checkout for the DUT), PORT (DUT serial), CLI_PORT (client serial), UPSTREAM (upstream bridge
+# address), SC (path to serial_capture.py), PY (checkout python)
+D=$(cd "$(dirname "$0")" && pwd); n=${1:?dut yaml name}; start=${2:?on or off}; dur=${3:?seconds}; shift 3
+T=${TREE:?set TREE}; P=${PORT:?set PORT}; CP=${CLI_PORT:?set CLI_PORT}; UP=${UPSTREAM:?set UPSTREAM}
+SC=${SC:?set SC}; PY=${PY:-$T/venv/bin/python}
+ap() { curl -s -m 5 -X POST -d "" -o /dev/null -w "%{http_code}" "http://$UP/button/ap%20$1/press"; }
+pkill -f "[s]erial_capture.py ($P|$CP)"
+[ "$(ap "$start")" = 200 ] || { echo "FAIL: upstream bridge did not take ap $start"; exit 1; }
+if [ "${CLI_FLASH:-0}" = 1 ]; then
+  (cd "$T" && PYTHONUNBUFFERED=1 timeout 1500 $PY -m esphome run --no-logs "$D/wispcli.yaml" --device "$CP" > "$D/wispcli-upload.log" 2>&1) \
+    || { echo "FAIL: client build or upload"; exit 1; }
+fi
+(cd "$T" && PYTHONUNBUFFERED=1 timeout 1500 $PY -m esphome run --no-logs "$D/$n.yaml" --device "$P" > "$D/$n-upload.log" 2>&1) \
+  || { echo "FAIL: DUT build or upload"; exit 1; }
+echo "$(date +%H:%M:%S) t=0 upstream $start start" > "$D/$n-marks.txt"  # phase 1 is the start state
+$PY "$SC" "$P" "$dur" "$D/$n-dut.txt" --reset > /dev/null 2>&1 & C1=$!
+# no --reset on the client: a CH340 D1 mini floods garbage after an RTS reset
+$PY "$SC" "$CP" "$dur" "$D/$n-cli.txt" > /dev/null 2>&1 & C2=$!
+trap 'kill $C1 $C2 2>/dev/null' EXIT
+t0=$(date +%s)
+for step in "$@"; do
+  t=${step%%:*}; st=${step#*:}
+  while [ $(($(date +%s) - t0)) -lt "$t" ]; do sleep 1; done
+  code=$(ap "$st")
+  echo "$(date +%H:%M:%S) t=$t upstream $st $code" | tee -a "$D/$n-marks.txt"
+  [ "$code" = 200 ] || { echo "FAIL: upstream bridge returned $code for ap $st"; exit 1; }
+done
+wait $C1 $C2; trap - EXIT
+$PY "$D/sum.py" "$D" "$n"

.agents/skills/hardware-test/scripts/wisp_ap/sum.py

@@ -0,0 +1,80 @@
+"""Summarise the client's results per upstream phase and the DUT's connects, AP starts, crashes and boots.
+Usage: sum.py DIR NAME   reads NAME-marks.txt, NAME-cli.txt and NAME-dut.txt
+Phase N runs from mark N (the first mark is the start state). Exits 1 when the DUT crashed, the DUT capture does not
+hold exactly one boot (empty or rebooted), the client had no successful result at all, or a phase with the upstream
+on had no successful name lookup (an answer equal to the gateway is the captive portal and does not count)."""
+
+from collections import defaultdict
+from pathlib import Path
+import re
+import sys
+
+D, n = Path(sys.argv[1]), sys.argv[2]
+
+
+def read(suffix: str) -> str:
+    return (D / f"{n}-{suffix}.txt").read_text(encoding="utf-8", errors="replace")
+
+
+mark_lines = [line.split() for line in read("marks").splitlines() if line.strip()]
+marks = [m[0] for m in mark_lines]
+upstream = {
+    i + 1: m[3] for i, m in enumerate(mark_lines)
+}  # "HH:MM:SS t=N upstream on|off CODE"
+cli = [line for line in read("cli").splitlines() if "[cli" in line]
+
+
+def phase(t: str) -> int:
+    return sum(1 for m in marks if t >= m)
+
+
+stat: dict[int, dict[str, int]] = defaultdict(lambda: defaultdict(int))
+good: dict[int, int] = defaultdict(int)  # successful name lookups per phase
+any_ok = 0
+for line in cli:
+    m = re.search(r"\]: (NOT CONNECTED|NET .*|NAME \S+|GW \S+|MAC \S+)", line)
+    if not m:
+        continue
+    s = m.group(1)
+    if s.startswith("NET"):
+        dns = re.search(r"dns (\S+)", s).group(1)
+        gw = re.search(r"gw (\S+)", s).group(1)
+        res = s.split("-> ")[1].strip()
+        # the captive portal answers every name with the AP's own address; that is not a working upstream
+        s = f"dns={dns} resolve={'CAPTIVE' if res == gw else res}"
+        if res not in ("FAIL", gw):
+            good[phase(line[1:9])] += 1
+            any_ok += 1
+    elif re.match(r"(NAME|GW|MAC) 200\b", s):
+        any_ok += 1
+    stat[phase(line[1:9])][s] += 1
+for p in sorted(stat):
+    print(f"phase {p} (upstream {upstream.get(p, '?')}):", dict(stat[p]))
+dead = [p for p, st in upstream.items() if st == "on" and not good[p]]
+dut = read("dut")
+crash = len(re.findall(r"Guru|panic|abort\(", dut))
+boots = dut.count("ESPHome version")
+print(
+    "DUT connected:",
+    len(re.findall(r"\[wifi:\d+\]: Connected", dut)),
+    "AP starts:",
+    dut.count("Starting AP"),
+    "crash:",
+    crash,
+    "boots:",
+    boots,
+)
+failed = []
+if not cli:
+    failed.append("empty client log")
+elif not any_ok:
+    failed.append("the client never got a successful result")
+if dead:
+    failed.append(f"no successful lookup in upstream on phases {dead}")
+# run.sh resets the DUT when the capture starts, so a good capture holds exactly one boot
+if boots != 1:
+    failed.append(f"{boots} DUT boots captured, expected 1")
+if crash:
+    failed.append(f"{crash} DUT crash lines")
+print("PASS" if not failed else "FAIL: " + "; ".join(failed))
+sys.exit(1 if failed else 0)

.agents/skills/hardware-test/scripts/wisp_ap/wisp-upstream.yaml

@@ -0,0 +1,52 @@
+esphome:
+  name: wisp-upstream
+
+esp32:
+  board: esp32-s3-devkitc-1
+  variant: esp32s3
+  flash_size: 16MB
+  framework:
+    type: esp-idf
+
+logger:
+  hardware_uart: USB_SERIAL_JTAG
+
+ethernet:
+  type: W5500
+  clk_pin: GPIO13
+  mosi_pin: GPIO11
+  miso_pin: GPIO12
+  cs_pin: GPIO14
+  interrupt_pin: GPIO10
+  reset_pin: GPIO9
+
+wifi:
+  ap:
+    ssid: !secret wisp_up_ssid
+    password: !secret wisp_up_password
+    ap_mode: WISP
+    manual_ip:
+      static_ip: 10.40.0.1
+      gateway: 10.40.0.1
+      subnet: 255.255.255.0
+
+api:
+  encryption:
+    key: !secret api_key
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+web_server:
+  port: 80
+
+button:
+  - platform: template
+    name: ap off
+    on_press:
+      - wifi.disable:
+  - platform: template
+    name: ap on
+    on_press:
+      - wifi.enable:

.agents/skills/hardware-test/scripts/wisp_ap/wispcli.yaml

@@ -0,0 +1,77 @@
+esphome:
+  name: wispcli
+
+esp8266:
+  board: d1_mini
+
+logger:
+  level: DEBUG
+
+wifi:
+  networks:
+    - ssid: !secret dut_ap_ssid
+      password: !secret dut_ap_password
+  reboot_timeout: 0s
+
+ota:
+  - platform: esphome
+
+http_request:
+  timeout: 5s
+  verify_ssl: false
+
+globals:
+  - id: tick
+    type: int
+    initial_value: "0"
+
+interval:
+  - interval: 3s
+    then:
+      - lambda: |-
+          id(tick)++;
+          if (id(tick) % 4 != 1) return;
+          if (!WiFi.isConnected()) { ESP_LOGI("cli", "NOT CONNECTED"); return; }
+          IPAddress r;
+          bool ok = WiFi.hostByName("neverssl.com", r, 3000);
+          ESP_LOGI("cli", "NET ip %s gw %s dns %s | neverssl.com -> %s", WiFi.localIP().toString().c_str(),
+                   WiFi.gatewayIP().toString().c_str(), WiFi.dnsIP(0).toString().c_str(),
+                   ok ? r.toString().c_str() : "FAIL");
+      - if:
+          condition:
+            lambda: return id(tick) % 4 == 2 && WiFi.isConnected();
+          then:
+            - http_request.get:
+                url: http://neverssl.com/
+                capture_response: true
+                max_response_buffer_size: 160
+                collect_headers:
+                  - location
+                on_response:
+                  - lambda: ESP_LOGI("cli", "NAME %d loc=%s body=%.60s", response->status_code, response->get_response_header("location").c_str(), body.c_str());
+                on_error:
+                  - lambda: ESP_LOGI("cli", "NAME error");
+      - if:
+          condition:
+            lambda: return id(tick) % 4 == 3 && WiFi.isConnected();
+          then:
+            - http_request.get:
+                url: !lambda 'return std::string("http://") + WiFi.gatewayIP().toString().c_str() + "/";'
+                capture_response: true
+                max_response_buffer_size: 160
+                on_response:
+                  - lambda: ESP_LOGI("cli", "GW %d body=%.60s", response->status_code, body.c_str());
+                on_error:
+                  - lambda: ESP_LOGI("cli", "GW error");
+      - if:
+          condition:
+            lambda: return id(tick) % 4 == 0 && WiFi.isConnected();
+          then:
+            - http_request.get:
+                url: !secret lan_probe_url
+                capture_response: true
+                max_response_buffer_size: 64
+                on_response:
+                  - lambda: ESP_LOGI("cli", "MAC %d body=%.20s", response->status_code, body.c_str());
+                on_error:
+                  - lambda: ESP_LOGI("cli", "MAC error");

.agents/skills/hardware-test/scripts/wpa3_ap/an.py

@@ -0,0 +1,56 @@
+"""Summarise the PMF AP's join and leave events per station and the BK7238's joins.
+Usage: an.py DIR BUILD [AP_SSID]   reads DIR/ap-BUILD.txt and DIR/bk-BUILD.txt (AP_SSID defaults to wd-pmf-ap)
+J = join, L<n> = leave with reason n. Reason 3: the station sent a deauth. Reason 2: the AP found a stale
+association when the station came back; a run of reason 2 before one join is reconnects the AP rejected.
+env DUT_MACS: comma separated station MACs that must appear in the AP events (default: at least two stations).
+Exits 1 on a rejected reconnect, an AP that did not switch to WPA2/WPA3, a missing DUT, or a BK7238 that never
+joined the AP."""
+
+from itertools import pairwise
+import os
+from pathlib import Path
+import re
+import sys
+
+D, b = Path(sys.argv[1]), sys.argv[2]
+AP_SSID = sys.argv[3] if len(sys.argv) > 3 else "wd-pmf-ap"
+failed = []
+ap = (D / f"ap-{b}.txt").read_text(encoding="utf-8", errors="replace").splitlines()
+if not any("PMF capable: 0" in line for line in ap) or any(
+    "AP CONFIG FAILED" in line for line in ap
+):
+    failed.append("the AP did not switch to WPA2/WPA3 mixed")
+ev = [line for line in ap if "STA JOIN" in line or "STA LEAVE" in line]
+mac_re = re.compile(r"([0-9A-F]{2}(?::[0-9A-F]{2}){5})")
+macs = sorted({m.group(1) for line in ev if (m := mac_re.search(line))})
+rejected = 0
+print(f"== {b}")
+for m in macs:
+    seq = [
+        "J" if "JOIN" in line else "L" + re.search(r"reason (\d+)", line).group(1)
+        for line in ev
+        if m in line
+    ]
+    # every L2 after the first in a run without a join in between is a rejected reconnect
+    rej = sum(1 for prev, cur in pairwise(seq) if prev == cur == "L2")
+    rejected += rej
+    print(" ", m, " ".join(seq), f"(rejected {rej})")
+if rejected:
+    failed.append(f"{rejected} rejected reconnects")
+if want := [
+    x.strip().upper() for x in os.environ.get("DUT_MACS", "").split(",") if x.strip()
+]:
+    if missing := [x for x in want if x not in macs]:
+        failed.append(f"no AP events for {', '.join(missing)}")
+elif len(macs) < 2:
+    failed.append(f"AP events for {len(macs)} stations, expected both DUTs")
+bk = (D / f"bk-{b}.txt").read_text(encoding="utf-8", errors="replace").splitlines()
+joins = [
+    m.groups() for line in bk if (m := re.search(r"JOINED (.+?) after (\d+) ms", line))
+]
+print("  bk7238 joins:", " ".join(f"{s}@{int(t) / 1000:.1f}s" for s, t in joins))
+print("  bk7238 reboots:", sum("REBOOT on" in line for line in bk))
+if not any(s == AP_SSID for s, _ in joins):
+    failed.append(f"the BK7238 never joined {AP_SSID}")
+print("PASS" if not failed else "FAIL: " + "; ".join(failed))
+sys.exit(1 if failed else 0)

.agents/skills/hardware-test/scripts/wpa3_ap/run.sh

@@ -0,0 +1,32 @@
+#!/bin/bash
+# usage: run.sh BUILD   flash the PMF AP, OTA the CB3S, serial flash the BK7238, then capture one AP window
+# Copy this directory into a scratch directory with a secrets.yaml (wifi_ssid, wifi_password, wd_ap_password) and
+# run it there. BUILD is a label (pr, base) for the output files.
+# env: TREE (esphome checkout for the DUTs), AP_PORT (classic devkit running wdap.yaml), BK_PORT (BK7238 serial),
+# CB_HOST (CB3S address on the house network), SC (path to serial_capture.py), FLASH_BK (path to flash_bk72xx.py),
+# LT (LibreTiny penv python, default ~/.platformio/penv/.libretiny/bin/python), PY (checkout python), DUT_MACS
+# (optional, comma separated station MACs of both DUTs for an.py)
+D=$(cd "$(dirname "$0")" && pwd); b=${1:?build label}; T=${TREE:?set TREE}; PY=${PY:-$T/venv/bin/python}
+AP=${AP_PORT:?set AP_PORT}; BK=${BK_PORT:?set BK_PORT}; CB=${CB_HOST:?set CB_HOST}; SC=${SC:?set SC}
+FB=${FLASH_BK:?set FLASH_BK}; LT=${LT:-$HOME/.platformio/penv/.libretiny/bin/python}
+pkill -f "[s]erial_capture.py ($AP|$BK)"
+# retry N PAUSE CMD...: run CMD up to N times, PAUSE seconds apart; fails when every attempt failed
+retry() { local n=$1 pause=$2 i; shift 2; for i in $(seq "$n"); do "$@" && return 0; [ "$i" -lt "$n" ] && sleep "$pause"; done; return 1; }
+build() { (cd "$T" && PYTHONUNBUFFERED=1 timeout 1500 $PY -m esphome "$@"); }
+build run --no-logs "$D/wdap.yaml" --device "$AP" > "$D/ap-$b-upload.log" 2>&1 || { echo "FAIL: AP build or upload"; exit 1; }
+# The DUT reboots 150 s after joining the house network; retry when it reboots mid-upload.
+cb_upload() { build run --no-logs "$D/wdcb.yaml" --device "$CB" > "$D/cb-$b-upload.log" 2>&1; }
+retry 3 20 cb_upload || { echo "FAIL: CB3S build or OTA after 3 tries"; exit 1; }
+build compile "$D/wdbk.yaml" > "$D/bk-$b-build.log" 2>&1 || { echo "FAIL: BK7238 build"; exit 1; }
+# The build directory has two .uf2 files; firmware.uf2 is the one to flash.
+UF2=$D/.esphome/build/wdbk/.pioenvs/wdbk/firmware.uf2
+bk_flash() { $LT "$FB" flash write -d "$BK" "$UF2" > "$D/bk-$b-flash.log" 2>&1; }
+retry 4 5 bk_flash || { echo "FAIL: BK7238 flash after 4 tries"; exit 1; }
+sleep 20
+# The reset restarts the AP's 420 s window.
+$PY "$SC" "$AP" 450 "$D/ap-$b.txt" --reset > /dev/null 2>&1 & C1=$!
+sleep 8
+$PY "$SC" "$BK" 440 "$D/bk-$b.txt" --released > /dev/null 2>&1 & C2=$!
+trap 'kill $C1 $C2 2>/dev/null' EXIT
+wait $C1 $C2; trap - EXIT
+$PY "$D/an.py" "$D" "$b"