@switchbot/openapi-cli 3.4.0 → 3.5.0

package/README.md

@@ -45,44 +45,14 @@ Under the hood every surface shares the same catalog, cache, and HMAC client —
 
 ## Table of contents
 
-- [Features](#features)
-- [Requirements](#requirements)
-- [Installation](#installation)
+- [Features](#features) · [Requirements](#requirements) · [Installation](#installation)
 - [Quick start](#quick-start)
 - [Credentials](#credentials)
-- [Policy](#policy)
+- [Policy](#policy) · [Rules engine](#rules-engine)
 - [Global options](#global-options)
-- [Commands](#commands)
-  - [`config`](#config--credential-management)
-  - [`devices`](#devices--list-status-control)
-  - [`devices batch`](#devices-batch--bulk-commands)
-  - [`devices watch`](#devices-watch--poll-status)
-  - [`scenes`](#scenes--run-manual-scenes)
-  - [`webhook`](#webhook--receive-device-events-over-http)
-  - [`events`](#events--receive-device-events)
-  - [`status-sync`](#status-sync--mqttopenclaw-bridge)
-  - [`plan`](#plan--declarative-batch-operations)
-  - [`mcp`](#mcp--model-context-protocol-server)
-  - [`doctor`](#doctor--self-check)
-  - [`health`](#health--runtime-health-report)
-  - [`upgrade-check`](#upgrade-check--version-check)
-  - [`quota`](#quota--api-request-counter)
-  - [`history`](#history--audit-log)
-  - [`catalog`](#catalog--device-type-catalog)
-  - [`schema`](#schema--export-catalog-as-json)
-  - [`capabilities`](#capabilities--cli-manifest)
-  - [`cache`](#cache--inspect-and-clear-local-cache)
-  - [`policy`](#policy--validate-scaffold-and-migrate-policyyaml)
-  - [`daemon`](#daemon--background-rules-engine-process)
-  - [`completion`](#completion--shell-tab-completion)
-- [Output modes](#output-modes)
-  - [Cache](#cache)
-- [Exit codes & error codes](#exit-codes--error-codes)
-- [Environment variables](#environment-variables)
-- [Scripting examples](#scripting-examples)
-- [Development](#development)
-- [License](#license)
-- [References](#references)
+- [Commands](#commands): [config](#config--credential-management) · [devices](#devices--list-status-control) · [scenes](#scenes--run-manual-scenes) · [webhook](#webhook--receive-device-events-over-http) · [events](#events--receive-device-events) · [status-sync](#status-sync--mqttopenclaw-bridge) · [daemon](#daemon--background-rules-engine-process) · [plan](#plan--declarative-batch-operations) · [mcp](#mcp--model-context-protocol-server) · [doctor](#doctor--self-check) · [health](#health--runtime-health-report) · [upgrade-check](#upgrade-check--version-check) · [quota](#quota--api-request-counter) · [history](#history--audit-log) · [catalog](#catalog--device-type-catalog) · [schema](#schema--export-catalog-as-json) · [capabilities](#capabilities--cli-manifest) · [cache](#cache--inspect-and-clear-local-cache) · [policy cmd](#policy--validate-scaffold-and-migrate-policyyaml) · [completion](#completion--shell-tab-completion)
+- [Output modes](#output-modes) · [Cache](#cache) · [Exit codes](#exit-codes--error-codes) · [Environment variables](#environment-variables)
+- [Scripting examples](#scripting-examples) · [Development](#development) · [License](#license)
 
 ---
 
@@ -93,7 +63,7 @@ Under the hood every surface shares the same catalog, cache, and HMAC client —
 - 🎨 **Dual output modes** — colorized tables by default; `--json` passthrough for `jq` and scripting
 - 🔐 **Secure credentials** — HMAC-SHA256 signed requests; config file written with `0600`; env-var override for CI
 - 🔍 **Dry-run mode** — preview every mutating request before it hits the API
-- 🧪 **Fully tested** — 2204 Vitest tests, mocked axios, zero network in CI
+- 🧪 **Fully tested** — 2263 Vitest tests, mocked axios, zero network in CI
 - ⚡ **Shell completion** — Bash / Zsh / Fish / PowerShell
 
 ## Requirements
@@ -296,110 +266,44 @@ then:
     template: '{"rule":"{{ rule.name }}","fired":"{{ rule.fired_at }}"}'
 ```
 
-**LLM condition** — add an AI judgement step before actions fire. The engine calls the
-configured LLM provider, passes the prompt plus recent event context, and gates execution
-on the model's yes/no answer:
+**LLM condition** — add an AI judgement step before actions fire:
 
 ```yaml
 conditions:
   - llm:
       prompt: "Is the temperature above normal comfort range?"
       provider: auto          # auto | openai | anthropic
-      cache_ttl: 5m           # skip redundant calls for identical context
+      cache_ttl: 5m
       budget:
         max_calls_per_hour: 20
       on_error: pass          # fail | pass | skip
 ```
 
-Set `OPENAI_API_KEY` or `ANTHROPIC_API_KEY` (provider `auto` tries Anthropic first).
-`rules lint` flags misconfigured LLM conditions (no provider key, cache TTL too high for
-the trigger frequency, budget zero). Evaluation decisions are recorded in the trace log.
+Set `OPENAI_API_KEY` or `ANTHROPIC_API_KEY`. `rules lint` flags misconfigured LLM conditions.
 
-**Decision trace** — enable `automation.audit.evaluate_trace` in `policy.yaml` to record
-every evaluation decision (why a rule fired or was blocked):
-
-```yaml
-automation:
-  audit:
-    evaluate_trace: sampled   # full | sampled | off (default: sampled)
-    evaluate_retention_days: 7
-```
+**Decision trace** — set `automation.audit.evaluate_trace: sampled` (or `full`) in `policy.yaml` to record every evaluation decision.
 
 ```bash
-# 1. Author rules under `automation.rules`. See examples/policies/automation.yaml
-#    for a walkthrough covering the three trigger sources.
-
-# 2. Static-check before running.
-switchbot rules lint                       # exit 0 valid, 1 error
-switchbot rules list --json | jq .         # structured summary
-
-# 3. Inspect a single rule in full detail (trigger, conditions, actions,
-#    cooldown, hysteresis, maxFiringsPerHour, suppressIfAlreadyDesired, last fired).
-switchbot rules explain "motion on"
-switchbot rules explain "motion on" --json
-
-# 4. Run the engine. --dry-run overrides every rule into audit-only mode;
-#    --max-firings bounds a demo session.
-switchbot rules run --dry-run --max-firings 5
-
-# 5. Edit policy.yaml in another shell, then hot-reload without restart.
-switchbot daemon reload                    # managed daemon reload
-
-# 6. Review recorded fires.
-switchbot rules tail --follow              # stream rule-* audit lines
-switchbot rules replay --since 1h --json   # per-rule fires/dries/throttled/errors
-switchbot rules summary                    # aggregate fires/errors per rule (24h window)
-switchbot rules last-fired -n 20           # 20 most recent fire entries
+switchbot rules lint                        # static check: exit 0 valid, 1 error
+switchbot rules list --json | jq .          # structured rule summary
+switchbot rules explain "motion on"         # trigger, conditions, actions, last fired
+switchbot rules run --dry-run --max-firings 5  # run engine; --dry-run = audit only
+switchbot daemon reload                     # hot-reload policy without restart
 
-# 7. Conflict and health analysis.
-switchbot rules conflicts                  # opposing actions, high-frequency MQTT,
-                                           # destructive commands, quiet-hours gaps
-switchbot rules doctor --json              # lint + conflicts combined; exit 0 when clean
+switchbot rules tail --follow               # stream rule-* audit lines
+switchbot rules replay --since 1h --json    # per-rule fires/dries/throttled/errors
+switchbot rules summary                     # aggregate fires/errors (24h)
+switchbot rules conflicts                   # opposing actions, destructive cmds, quiet-hours gaps
+switchbot rules doctor --json               # lint + conflicts; exit 0 when clean
 
-# 8. Scaffold a new rule from natural language (heuristic or LLM-backed).
 switchbot rules suggest --intent "turn off AC at 11pm"
-switchbot rules suggest --intent "if door opens and temp below 20 turn on heater" \
-  --llm auto                               # routes complex intents to LLM automatically
-switchbot rules suggest --intent "..." --llm openai  # explicit backend
-# Set OPENAI_API_KEY or ANTHROPIC_API_KEY; auto mode falls back to heuristic on failure
-
-# 9. Explain why a specific evaluation fired or was blocked (requires evaluate_trace).
-switchbot rules trace-explain --rule "motion on" --last
-switchbot rules trace-explain --rule "motion on" --since 1h --json
-switchbot rules trace-explain <fireId>     # single evaluation by ID
-
-# 10. Simulate a rule against historical events without running the engine.
-switchbot rules simulate "motion on"       # replay last 24h from audit log
-switchbot rules simulate "motion on" --since 7d --json
-switchbot rules simulate policy.yaml --rule "night AC" --against events.jsonl
-```
-
-`rules suggest` enforces several guardrails on LLM output so a model can't quietly arm
-something unsafe:
-
-- **`dry_run` is forced to `true`** on every LLM-generated rule. Review the output and
-  flip it yourself before running the engine without `--dry-run`.
-- **Explicit overrides always win.** If you pass `--trigger`, the LLM's answer must match;
-  a mismatch fails fast. Within the same trigger, mismatched `--event` / `--schedule` /
-  `--days` / `--webhook-path` are rewritten to your value with a warning.
-- **`--llm` is enum-validated at the CLI** (`auto | openai | anthropic`) — junk values
-  exit non-zero instead of falling through.
-- **Notify URLs must be `http://` or `https://`.** `rules lint` and the runtime both
-  reject `file://`, `ftp://`, etc., so a generated webhook can't smuggle in a non-HTTP
-  scheme.
-
-When `quiet_hours` is configured in `policy.yaml`, `rules conflicts` additionally flags event-driven (MQTT / webhook) rules that lack a `time_between` condition — they would fire uninhibited during the quiet window. The hint in each finding includes a ready-to-paste `time_between` condition to add.
+switchbot rules suggest --intent "..." --llm auto    # LLM-backed (OPENAI_API_KEY or ANTHROPIC_API_KEY)
 
-Webhook trigger token management:
-
-```bash
-switchbot rules webhook-rotate-token       # rotate the bearer token for webhook triggers
-switchbot rules webhook-show-token         # print current token (creates one if absent)
+switchbot rules trace-explain --rule "motion on" --last   # why a rule fired/was blocked
+switchbot rules simulate "motion on" --since 7d --json    # replay without running the engine
 ```
 
-See [`docs/design/phase4-rules.md`](./docs/design/phase4-rules.md) for
-the engine's pipeline (subscribe → classify → match → conditions →
-throttle → action → audit).
+LLM-generated rules always have `dry_run: true` — flip it yourself after review. Notify URLs must be `http://` or `https://`. See [`docs/design/phase4-rules.md`](./docs/design/phase4-rules.md) for the full pipeline.
 
 ## Global options
 
@@ -424,23 +328,11 @@ throttle → action → audit).
 - `-V`, `--version`: Print the CLI version.
 - `-h`, `--help`: Show help for any command or subcommand.
 
-Every subcommand supports `--help`, and most include a parameter-format reference and examples.
-
-```bash
-switchbot --help
-switchbot devices command --help
-```
-
-> **Tip — required-value flags and subcommands.** Flags like `--profile`, `--timeout`, `--max`, and `--interval` take a value. If you omit it, Commander will happily consume the next token — including a subcommand name. Since v2.2.1 the CLI rejects that eagerly (exit 2 with a clear error), but if you ever hit `unknown command 'list'` after something like `switchbot --profile list`, use the `--flag=value` form: `switchbot --profile=home devices list`.
+Every subcommand supports `--help`. Use `--flag=value` form when a flag takes a value and is followed by a subcommand (e.g. `switchbot --profile=home devices list`).
 
 ### `--dry-run`
 
-Intercepts every non-GET request: the CLI prints the URL/body it would have
-sent, then exits `0` without contacting the API. `GET` requests (list, status,
-query) are still executed so you can preview the state involved. Dry-run also
-validates command names against the device catalog and rejects unknown commands
-(exit 2) when the device type has a known catalog entry. Commands sent to
-read-only sensors (e.g. Meter) are likewise rejected.
+Intercepts every non-GET request: prints the URL/body it would have sent, then exits `0`. GET requests still execute. Also validates command names against the device catalog (exit 2 on unknown commands or read-only sensors).
 
 ```bash
 switchbot devices command ABC123 turnOn --dry-run
@@ -456,127 +348,44 @@ switchbot devices command ABC123 turnOn --dry-run
 switchbot config set-token <token> <secret>   # Save to ~/.switchbot/config.json
 switchbot config show                          # Print current source + masked secret
 switchbot config list-profiles                 # List saved profiles
-
-# Print (or write) the recommended AI-agent profile template
-switchbot config agent-profile                 # print to stdout
-switchbot config agent-profile --write         # write to ~/.switchbot/profiles/agent.json (mode 0600)
-switchbot config agent-profile --write --force # overwrite if it already exists
-switchbot config agent-profile --json          # structured JSON envelope
+switchbot config agent-profile --write         # write recommended AI-agent profile (mode 0600)
 ```
 
 ### `devices` — list, status, control
 
 ```bash
 # List all physical devices and IR remote devices
-# Default columns (4): deviceId, deviceName, type, category
-# Pass --wide for the full 10-column operator view
-switchbot devices list
-switchbot devices ls              # short alias for 'list'
-switchbot devices list --wide
+switchbot devices list                         # default 4 columns: deviceId, deviceName, type, category
+switchbot devices list --wide                  # full 10-column operator view
 switchbot devices list --json | jq '.deviceList[].deviceId'
-
-# IR remotes: type = remoteType (e.g. "TV"), category = "ir"
-# Physical: category = "physical"
 switchbot devices list --format=tsv --fields=deviceId,type,category
 
-# Filter devices by type / name / category / room (server-side filter keys)
-switchbot devices list --filter category=physical
-switchbot devices list --filter type=Bot
-switchbot devices list --filter name=living,category=physical
-
-# Filter operators: = (substring; exact for `category`), ~ (substring),
-# =/regex/ (case-insensitive regex). Clauses are AND-ed.
-switchbot devices list --filter 'name~living'
-switchbot devices list --filter 'type=/Hub.*/'
-switchbot devices list --filter 'name~office,type=/Bulb|Strip/'
-
-# Filter by family / room (family & room info requires the platform source
-# header, which this CLI sends on every request)
-switchbot devices list --json | jq '.deviceList[] | select(.familyName == "Home")'
-switchbot devices list --json | jq '[.deviceList[], .infraredRemoteList[]] | group_by(.familyName)'
+# Filter by type / name / category / room
+# Operators: = (substring; exact for category), ~ (substring), =/regex/; clauses AND-ed
+switchbot devices list --filter 'type=Bot'
+switchbot devices list --filter 'name~living,type=/Bulb|Strip/'
+switchbot devices list --filter 'category=physical'
 
-# Query real-time status of a physical device
+# Query real-time status
 switchbot devices status <deviceId>
-switchbot devices status <deviceId> --json
+switchbot devices status --ids ABC,DEF,GHI    # batch status
+switchbot devices status --ids ABC,DEF --fields power,battery --format jsonl
 
 # Resolve device by fuzzy name instead of ID (status, command, describe, expand, watch)
 switchbot devices status --name "Living Room AC"
 switchbot devices command --name "Office Light" turnOn
-switchbot devices describe --name "Kitchen Bot"
-
-# Batch status across multiple devices
-switchbot devices status --ids ABC,DEF,GHI
-switchbot devices status --ids ABC,DEF --fields power,battery  # only show specific fields
-switchbot devices status --ids ABC,DEF --format jsonl           # one JSON line per device
 
 # Send a control command
 switchbot devices command <deviceId> <cmd> [parameter] [--type command|customize]
 
-# Describe a specific device (1 API call): metadata + supported commands + status fields
-switchbot devices describe <deviceId>
-switchbot devices describe <deviceId> --json
-
-# Discover what's supported (offline reference, no API call)
-switchbot devices types                 # List all device types + IR remote types (incl. role column)
-switchbot devices commands <type>       # Show commands, parameter formats, and status fields
-switchbot devices commands Bot
-switchbot devices commands "Smart Lock"
-switchbot devices commands curtain      # Case-insensitive, substring match
+# Offline reference (no API call)
+switchbot devices types                        # all device types
+switchbot devices commands <type>              # commands, parameter formats, status fields
 ```
 
-#### Filter expressions — per-command reference
-
-Three commands accept `--filter`. They share one four-operator grammar,
-but each exposes its own key set:
-
-- `devices list`
-  Operators: `=` (substring; **exact** for `category`), `!=` (negated),
-  `~` (substring), `=/regex/` (case-insensitive regex).
-  Keys: `type`, `name`, `category`, `room`.
-- `devices batch`
-  Operators: same as `devices list`.
-  Keys: `type`, `family`, `room`, `category`.
-- `events tail` / `events mqtt-tail`
-  Operators: same (tail only; mqtt-tail uses `--topic` instead).
-  Keys: `deviceId`, `type`.
+Parameters for `setAll`, `setPosition`, `setMode`, `setBrightness`, and `setColor` are validated client-side (exit 2 on bad input). `setColor` accepts `R:G:B`, `#RRGGBB`, `#RGB`, and CSS names — all normalize to `R:G:B`. Pass `--skip-param-validation` to bypass. Unknown deviceIds exit 2 by default; pass `--allow-unknown-device` for scripted pass-through.
 
-Clauses are comma-separated and AND-ed. No OR across clauses — use regex
-alternation (`=/A|B/`) for that. `category` is the one key that stays exact
-under `=` / `!=` to preserve `category=physical` / `category!=ir` semantics.
-A clause with an empty value (e.g. `name~`, `type=`) is rejected with exit 2 —
-the parser refuses to guess whether an empty value means "no constraint" or
-"match empty string". Drop the clause outright to remove the constraint.
-
-#### Parameter formats
-
-`parameter` is optional — omit it for commands like `turnOn`/`turnOff` (auto-defaults to `"default"`).
-Numeric-only and JSON-object parameters are auto-parsed; strings with colons / commas / semicolons pass through as-is.
-
-For the exact commands and parameter formats a specific device supports, query the built-in catalog:
-
-```bash
-switchbot devices commands <type>       # e.g. Bot, Curtain, "Smart Lock", "Robot Vacuum Cleaner S10"
-```
-
-Generic parameter shapes (which one applies is decided by the device — see the catalog):
-
-| Shape               | Example                                                  |
-| ------------------- | -------------------------------------------------------- |
-| _(none)_            | `devices command <id> turnOn`                            |
-| `<integer>`         | `devices command <id> setBrightness 75`                  |
-| `<R:G:B>`           | `devices command <id> setColor "255:0:0"`                |
-| `<direction;angle>` | `devices command <id> setPosition "up;60"`               |
-| `<a,b,c,…>`         | `devices command <id> setAll "26,1,3,on"`                |
-| `<json object>`     | `'{"action":"sweep","param":{"fanLevel":2,"times":1}}'`  |
-| Custom IR button    | `devices command <id> MyButton --type customize`         |
-
-Parameters for `setAll` (Air Conditioner), `setPosition` (Curtain / Blind Tilt), `setMode` (Relay Switch), `setBrightness` (dimmable lights), and `setColor` (Color Bulb / Strip Light / Ceiling Light) are validated client-side before the request — malformed shapes, out-of-range values, and JSON for CSV fields all fail fast with exit 2. `setColor` accepts `R:G:B`, `R,G,B`, `#RRGGBB`, `#RGB`, and CSS named colors (`red`, `blue`, …); all normalize to `R:G:B` before hitting the API. Pass `--skip-param-validation` to bypass (escape hatch — prefer fixing the argument). Command names are also case-normalized against the catalog (e.g. `turnon` is auto-corrected to `turnOn` with a stderr warning); unknown names still exit 2 with the supported-commands list.
-
-Unknown deviceIds (not in the local cache) exit 2 by default so `--dry-run` is a reliable pre-flight gate. Unknown command names and commands on read-only sensors are also rejected during dry-run when the device type has a catalog entry. Run `switchbot devices list` first, or pass `--allow-unknown-device` for scripted pass-through.
-
-Negative numeric parameters (e.g. `setBrightness -1` for a probe) are passed through to the command validator instead of being swallowed by the flag parser as an unknown option.
-
-For the complete per-device command reference, see the [SwitchBot API docs](https://github.com/OpenWonderLabs/SwitchBotAPI#send-device-control-commands).
+For per-device command and parameter details: `switchbot devices commands <type>` or the [SwitchBot API docs](https://github.com/OpenWonderLabs/SwitchBotAPI#send-device-control-commands).
 
 #### `devices expand` — named flags for packed parameters
 
@@ -596,61 +405,46 @@ switchbot devices expand <blindId> setPosition --direction up --angle 50
 
 # Relay Switch — setMode
 switchbot devices expand <relayId> setMode --channel 1 --mode edge
+
+# Color Bulb / Strip Light / Floor Lamp / Ceiling Light — setBrightness / setColor / setColorTemperature
+switchbot devices expand <bulbId> setBrightness --brightness 80
+switchbot devices expand <bulbId> setColor --color "#FF0000"
+switchbot devices expand <bulbId> setColorTemperature --color-temp 4000
 ```
 
-Run `switchbot devices expand <id> <command> --help` to see the available flags for any device command. `expand` is only meaningful for multi-parameter commands (the four above); single-parameter commands like `setBrightness 50` or `setColor "#FF0000"` are already flag-free at the CLI level.
+Run `switchbot devices expand <id> <command> --help` to see the available flags for any device command.
 
 #### `devices explain` — one-shot device summary
 
 ```bash
-# Metadata + supported commands + live status in one call
-switchbot devices explain <deviceId>
-
-# Skip live status fetch (catalog-only output, no API call)
-switchbot devices explain <deviceId> --no-live
+switchbot devices explain <deviceId>          # metadata + commands + live status
+switchbot devices explain <deviceId> --no-live # catalog-only, no API call
 ```
 
-Returns a combined view: static catalog info (commands, parameters, status fields) merged with the current live status. For Hub devices, also lists connected child devices. Prefer this over separate `status` + `describe` calls.
-
 #### `devices meta` — local device metadata
 
 ```bash
 switchbot devices meta set <deviceId> --alias "Office Light"
-switchbot devices meta set <deviceId> --hide          # hide from `devices list`
+switchbot devices meta set <deviceId> --hide   # hide from `devices list`
 switchbot devices meta get <deviceId>
-switchbot devices meta list                            # show all saved metadata
+switchbot devices meta list
 switchbot devices meta clear <deviceId>
 ```
 
-Stores local annotations (alias, hidden flag, notes) in `~/.switchbot/device-meta.json`. The alias is used as a display name; `--show-hidden` on `devices list` reveals hidden devices.
+Stores local annotations in `~/.switchbot/device-meta.json`. `--show-hidden` on `devices list` reveals hidden devices.
 
 #### `devices batch` — bulk commands
 
 ```bash
-# Send the same command to every device matching a filter
+# Same command to every matching device
 switchbot devices batch turnOff --filter 'type=Bot'
 switchbot devices batch setBrightness 50 --filter 'type~Light,family=Living'
-
-# Explicit device IDs (comma-separated)
 switchbot devices batch turnOn --ids ID1,ID2,ID3
-
-# Pipe device IDs from `devices list`
 switchbot devices list --format=id --filter 'type=Bot' | switchbot devices batch toggle -
-
-# Destructive commands require --yes
-switchbot devices batch unlock --filter 'type=Smart Lock' --yes
-
-# Skip devices whose cached status is offline (default: off)
-switchbot devices batch turnOn --ids ID1,ID2 --skip-offline
-
-# --idempotency-key is an alias for --idempotency-key-prefix; both append -<deviceId>
-switchbot devices batch turnOn --ids ID1,ID2 --idempotency-key morning-lights
+switchbot devices batch unlock --filter 'type=Smart Lock' --yes  # destructive: requires --yes
 ```
 
-Sends the same command to many devices in one run. Filter grammar matches `devices list` (`=` substring, `~` substring, `=/regex/` regex — clauses AND-ed); supported keys here are `type`, `family`, `room`, `category`. Destructive commands (Smart Lock unlock, Garage Door Opener, etc.) require `--yes` to prevent accidents.
-
-`--skip-offline` reads from the local status cache only (no new API calls);
-skipped devices appear under `summary.skipped` with `skippedReason:'offline'`.
+Filter keys: `type`, `family`, `room`, `category`. Skipped-offline devices appear under `summary.skipped` when `--skip-offline` is passed.
 
 ### `scenes` — run manual scenes
 
@@ -685,140 +479,39 @@ The CLI validates that `<url>` is an absolute `http://` or `https://` URL before
 
 ### `events` — receive device events
 
-Two subcommands cover the two ways SwitchBot can push state changes to you.
-
 #### `events tail` — local webhook receiver
 
 ```bash
-# Listen on port 3000 and print every incoming webhook POST
-switchbot events tail
-
-# Filter to one device
-switchbot events tail --filter deviceId=ABC123
-
-# Stop after 5 matching events
-switchbot events tail --filter 'type=WoMeter' --max 5
-
-# Stop after 10 minutes regardless of event count
-switchbot events tail --for 10m
-
-# Custom port / path
+switchbot events tail                              # listen on port 3000
+switchbot events tail --filter deviceId=ABC123     # filter to one device
+switchbot events tail --filter 'type=WoMeter' --max 5 --for 10m
 switchbot events tail --port 8080 --path /hook --json
 ```
 
-Run `switchbot webhook setup https://your.host/hook` first to tell SwitchBot where to send events, then expose the local port via ngrok/cloudflared and point the webhook URL at it. `events tail` only runs the local receiver — tunnelling is up to you.
-
-Output (one JSON line per matched event):
-
-```json
-{ "t": "2024-01-01T12:00:00.000Z", "remote": "1.2.3.4:54321", "path": "/", "body": {...}, "matched": true }
-```
-
-Filter keys: `deviceId`, `type`. Operators: `=` (substring), `~` (substring), `=/regex/` (case-insensitive regex). Clauses comma-separated and AND-ed.
+Run `switchbot webhook setup https://your.host/hook` first. `events tail` only runs the local receiver — tunnelling (ngrok/cloudflared) is up to you.
 
 #### `events mqtt-tail` — real-time MQTT stream
 
 ```bash
-# Stream all shadow-update events (runs in foreground until Ctrl-C)
-switchbot events mqtt-tail
-
-# Filter to a topic subtree
-switchbot events mqtt-tail --topic 'switchbot/#'
-
-# Stop after 10 events
-switchbot events mqtt-tail --max 10 --json
-
-# Stop after a fixed duration (emits __session_start under --json before connect)
-switchbot events mqtt-tail --for 30s --json
+switchbot events mqtt-tail                         # stream all shadow events (Ctrl-C to stop)
+switchbot events mqtt-tail --topic 'switchbot/#'   # filter to topic subtree
+switchbot events mqtt-tail --max 10 --for 30s --json
 ```
 
-Connects to the SwitchBot MQTT service automatically using the same credentials configured for the REST API (`SWITCHBOT_TOKEN` + `SWITCHBOT_SECRET`). No additional MQTT configuration is required — the client certificates are provisioned on first use.
-
-Output (one JSON line per message):
-
-```json
-{ "t": "2024-01-01T12:00:00.000Z", "topic": "switchbot/abc123/status", "payload": {...} }
-```
-
-This command runs in the foreground and streams events until you press Ctrl-C. To run it persistently in the background, use a process manager:
-
-```bash
-# pm2
-pm2 start "switchbot events mqtt-tail --json" --name switchbot-events
-
-# nohup
-nohup switchbot events mqtt-tail --json >> ~/switchbot-events.log 2>&1 &
-```
-
-Run `switchbot doctor` to verify MQTT credentials are configured correctly before connecting.
+Credentials are provisioned automatically from the REST API config. Use `--sink` to route events to external services (`file`, `webhook`, `telegram`, `homeassistant`, `openclaw`) — see `switchbot events mqtt-tail --help` for details.
 
 ### `status-sync` — MQTT/OpenClaw bridge
 
-Use this command family when you want the CLI itself to own the lifecycle of a
-long-running bridge that forwards SwitchBot MQTT shadow events into an OpenClaw
-gateway. Internally it reuses `events mqtt-tail --sink openclaw`, but adds a
-stable command surface for foreground execution, background startup, status
-inspection, and shutdown.
+Forwards SwitchBot MQTT shadow events into an OpenClaw gateway with stable lifecycle management.
 
 ```bash
-# Foreground mode for supervisors / containers
-switchbot status-sync run --openclaw-model home-agent
-
-# Background mode for a normal shell session
-switchbot status-sync start --openclaw-model home-agent
-
-# Inspect the current bridge
+switchbot status-sync run --openclaw-model home-agent   # foreground (for supervisors)
+switchbot status-sync start --openclaw-model home-agent # background
 switchbot status-sync status --json
-
-# Stop the running bridge
 switchbot status-sync stop
 ```
 
-Required input:
-
-- `OPENCLAW_MODEL` or `--openclaw-model <id>`
-- `OPENCLAW_TOKEN` or `--openclaw-token <token>`
-
-Optional input:
-
-- `OPENCLAW_URL` or `--openclaw-url <url>`
-- `--topic <pattern>` to narrow the MQTT subscription
-- `SWITCHBOT_STATUS_SYNC_HOME` or `--state-dir <path>` for custom runtime state
-
-Background mode writes these files under the state directory:
-
-- `state.json` — current pid, start time, effective command
-- `stdout.log` — child stdout
-- `stderr.log` — child stderr
-
-Foreground vs background:
-
-- `status-sync run` keeps the bridge attached to the current terminal
-- `status-sync start` detaches the bridge and returns immediately
-- `status-sync status` reports whether the bridge is alive plus paths/logs
-- `status-sync stop` terminates the managed bridge process tree
-
-#### `mqtt-tail` sinks — route events to external services
-
-By default `mqtt-tail` prints JSONL to stdout. Use `--sink` (repeatable) to route events to one or more destinations instead:
-
-| Sink | Required flags |
-| --- | --- |
-| `stdout` | (default when no `--sink` given) |
-| `file` | `--sink-file <path>` — append JSONL |
-| `webhook` | `--webhook-url <url>` — HTTP POST each event |
-| `telegram` | `--telegram-token` (or `$TELEGRAM_TOKEN`), `--telegram-chat <chatId>` |
-| `homeassistant` | `--ha-url <url>` + `--ha-webhook-id` (no auth) or `--ha-token` (REST event API) |
-
-```bash
-# Generic webhook (n8n, Make, etc.)
-switchbot events mqtt-tail --sink webhook --webhook-url https://n8n.local/hook/abc
-
-# Forward to Home Assistant via webhook trigger
-switchbot events mqtt-tail --sink homeassistant --ha-url http://homeassistant.local:8123 --ha-webhook-id switchbot
-```
-
-Device state is also persisted to `~/.switchbot/device-history/<deviceId>.json` (latest + 100-entry ring buffer) regardless of sink configuration. This enables the `get_device_history` MCP tool to answer state queries without an API call.
+Required: `OPENCLAW_MODEL` (or `--openclaw-model`) and `OPENCLAW_TOKEN`. Optional: `OPENCLAW_URL`, `--topic`, `--state-dir`. Background mode writes `state.json`, `stdout.log`, and `stderr.log` under the state directory.
 
 ### `daemon` — background rules-engine process
 
@@ -917,10 +610,11 @@ switchbot mcp serve
 ```
 
 Exposes MCP tools (`list_devices`, `describe_device`, `get_device_status`,
+`get_device_history`, `query_device_history`, `aggregate_device_history`,
 `send_command`, `list_scenes`, `run_scene`, `search_catalog`,
 `account_overview`, `plan_suggest`, `plan_run`, `audit_query`,
 `audit_stats`, `policy_diff`, `policy_validate`, `policy_new`,
-`policy_migrate`, `rules_suggest`, `rule_notifications`,
+`policy_migrate`, `policy_add_rule`, `rules_suggest`, `rule_notifications`,
 `rules_explain`, `rules_simulate`) plus a
 `switchbot://events` resource for real-time shadow updates.
 `rules_suggest` accepts an optional `llm` parameter (`openai | anthropic | auto`)
@@ -939,7 +633,7 @@ switchbot doctor
 switchbot doctor --json
 ```
 
-Runs local checks (Node version, credentials, profiles, catalog, cache, quota, clock, MQTT, policy, MCP, notify-connectivity) and exits 1 if any check fails. `warn` results exit 0. The MQTT check reports `ok` when REST credentials are configured (auto-provisioned on first use). The `notify-connectivity` check probes webhook URLs declared in `type: notify` actions. Use this to diagnose connectivity or config issues before running automation.
+Runs local checks (Node version, credentials, profiles, catalog, catalog-schema, cache, quota, clock, MQTT, policy, MCP, keychain, path, inventory, audit, daemon, health, notify-connectivity, release-notes) and exits 1 if any check fails. `warn` results exit 0. The MQTT check reports `ok` when REST credentials are configured (auto-provisioned on first use). The `notify-connectivity` check probes webhook URLs declared in `type: notify` actions. Use this to diagnose connectivity or config issues before running automation.
 
 `--json` output includes `maturityScore` (0–100) and `maturityLabel` (`production-ready` / `mostly-ready` / `needs-work` / `not-ready`) to give an at-a-glance readiness rating:
 
@@ -970,71 +664,44 @@ Port conflicts are reported immediately with a clear hint to choose a different
 ### `upgrade-check` — version check
 
 ```bash
-switchbot upgrade-check                      # human output; exits 1 when update available
-switchbot upgrade-check --json               # structured JSON output
-switchbot upgrade-check --timeout 5000       # custom registry timeout (ms)
-```
-
-Queries the npm registry for the latest published version and compares it against the running version. When the registry's `dist-tags.latest` is itself a prerelease (e.g. `4.0.0-rc.1`), the check is skipped and the current version is treated as up-to-date — accidental prerelease tags don't trigger spurious upgrade prompts.
-`--json` output:
-
-```json
-{
-  "current": "3.3.2",
-  "latest": "4.0.0",
-  "upToDate": false,
-  "updateAvailable": true,
-  "breakingChange": true,
-  "installCommand": "npm install -g @switchbot/openapi-cli@4.0.0"
-}
+switchbot upgrade-check        # exits 1 when update available
+switchbot upgrade-check --json # {current, latest, upToDate, updateAvailable, breakingChange, installCommand}
 ```
 
-`breakingChange` is `true` when the latest major version is higher than the current — useful for agents or CI that need to distinguish breaking upgrades from patch releases.
-
 ### `quota` — API request counter
 
 ```bash
-switchbot quota status     # today's usage + last 7 days
-switchbot quota reset      # delete the counter file
+switchbot quota status     # today's usage + last 7 days (10,000/day limit)
+switchbot quota reset
 ```
 
-Tracks daily API calls against the 10,000/day account limit. The counter is stored in `~/.switchbot/quota.json` and incremented on every mutating request. Pass `--no-quota` to skip tracking for a single run.
-
 ### `history` — audit log
 
 ```bash
-switchbot history show              # recent entries (newest first)
-switchbot history show --limit 20   # last 20 entries
-switchbot history replay 7          # re-run entry #7
+switchbot history show --limit 20
+switchbot history replay 7           # re-run entry #7
 switchbot --json history show --limit 50 | jq '.entries[] | select(.result=="error")'
 ```
 
-Reads the JSONL audit log (`~/.switchbot/audit.log` by default; override with `--audit-log --audit-log-path <path>`). Each entry records the timestamp, command, device ID, result, and dry-run flag. `replay` re-runs the original command with the original arguments.
-
 ### `catalog` — device type catalog
 
 ```bash
-switchbot catalog show              # all 42 built-in types
-switchbot catalog list              # alias for `show`
+switchbot catalog show              # all built-in types
 switchbot catalog show Bot          # one type
-switchbot catalog search Hub        # fuzzy match across type / aliases / commands
-switchbot catalog diff              # what a local overlay changes vs built-in
-switchbot catalog path              # location of the local overlay file
-switchbot catalog refresh           # reload local overlay (clears in-process cache)
+switchbot catalog search Hub        # fuzzy match
+switchbot catalog diff              # local overlay vs built-in
 ```
 
-The built-in catalog ships with the package. Create `~/.switchbot/catalog-overlay.json` to add, extend, or override type definitions without modifying the package.
+Create `~/.switchbot/catalog-overlay.json` to extend or override type definitions without modifying the package.
 
 ### `schema` — export catalog as JSON
 
 ```bash
-switchbot schema export                         # all types as structured JSON
-switchbot schema export --type 'Strip Light'    # one type
-switchbot schema export --role sensor           # filter by role
+switchbot schema export                         # all types
+switchbot schema export --type 'Strip Light'
+switchbot schema export --role sensor
 ```
 
-Exports the effective catalog in a machine-readable format. Pipe the output into an agent's system prompt or tool schema to give it a complete picture of controllable devices.
-
 ### `capabilities` — CLI manifest
 
 ```bash
@@ -1042,127 +709,58 @@ switchbot capabilities --json
 switchbot capabilities --used --json   # only types seen in the local cache
 ```
 
-Prints a versioned JSON manifest describing available surfaces (CLI, MCP, MQTT, plan runner), commands, and environment variables. Every subcommand leaf now carries a `{mutating, consumesQuota, idempotencySupported, agentSafetyTier, verifiability, typicalLatencyMs}` block, and the top-level payload publishes a flat `commandMeta` path-keyed lookup so agents don't have to walk the tree. `--used` filters the per-type summary to devices actually present in the local cache (same semantics as `schema export --used`).
+Prints a versioned manifest of surfaces, commands, and environment variables. Each command leaf includes `{mutating, consumesQuota, agentSafetyTier, typicalLatencyMs}`.
 
 ### `cache` — inspect and clear local cache
 
 ```bash
-# Show cache status (paths, age, entry counts)
-switchbot cache show
-
-# Clear everything
-switchbot cache clear
-
-# Clear only the device-list cache or only the status cache
-switchbot cache clear --key list
-switchbot cache clear --key status
+switchbot cache show                   # paths, age, entry counts
+switchbot cache clear                  # clear everything
+switchbot cache clear --key list       # list cache only
+switchbot cache clear --key status     # status cache only
 ```
 
 ### `policy` — validate, scaffold, and migrate policy.yaml
 
-Companion to the separate SwitchBot skill repository for third-party agent hosts. The skill reads behaviour (aliases, confirmations, quiet hours, audit path) from `policy.yaml`. This command group checks that file before the skill ever sees it, turning what used to be silent failures into line-accurate errors.
-
 ```bash
-# Write a starter policy at the default location
-switchbot policy new                              # writes to the resolved default policy path
-switchbot policy new ./custom/policy.yaml --force
-
-# Validate (compiler-style errors with line:col + caret + hints)
-switchbot policy validate
-switchbot policy validate ./custom/policy.yaml
+switchbot policy new                              # write a starter policy
+switchbot policy validate                         # compiler-style errors (line:col + caret)
 switchbot policy validate --json | jq '.data.errors'
-switchbot policy validate --no-snippet             # plain error list, no source preview
-
-# Report the schema version the file declares
-switchbot policy migrate
-
-# Snapshot and restore the active policy
-switchbot policy backup                            # write timestamped backup alongside policy file
-switchbot policy backup --out ./backups/           # custom destination directory
-switchbot policy restore <backup-file>             # overwrite active policy from backup (auto-backups first)
+switchbot policy migrate                          # upgrade v0.1 → v0.2 in-place
+switchbot policy backup                           # timestamped backup
+switchbot policy restore <backup-file>
 ```
 
-Path resolution order: positional `[path]` > `SWITCHBOT_POLICY_PATH` env var > default policy path.
-
-**Exit codes:** `0` valid / `1` invalid / `2` file-not-found / `3` yaml-parse / `4` internal / `5` file already exists (on `new`, overridden with `--force`) / `6` unsupported schema version (on `migrate`).
-
-Example — editing an alias without quoting the deviceId:
-
-```console
-$ switchbot policy validate
-<policy-path>:14:11
-  14 |   bedroom light: 01-abc-12345
-                 ^^^^^^^^^^^^^
-error: /aliases/bedroom light does not match pattern ^[A-Z0-9]{2,}-[A-Z0-9-]+$
-hint:  paste the deviceId from `switchbot devices list --format=tsv`, e.g. 01-202407090924-26354212
-
-✗ 1 error in <policy-path> (schema v0.1)
-```
-
-The default policy schema shipped with the CLI (`src/policy/schema/v0.2.json`) is mirrored as `examples/policy.schema.json` in the companion skill repo; a CI job on every push diffs the two to prevent drift.
+Path resolution: positional `[path]` > `SWITCHBOT_POLICY_PATH` > default. Exit codes: `0` valid / `1` invalid / `2` missing / `3` yaml-parse / `4` internal / `5` exists (use `--force`) / `6` unsupported version.
 
 ## Output modes
 
 - **Default** — ANSI-colored tables for `list`/`status`, key-value tables for details.
-- **`--json`** — raw API payload passthrough. Output is the exact JSON the SwitchBot API returned, ideal for `jq` and scripting. Errors are also JSON on stderr: `{ "error": { "code", "kind", "message", "hint?" } }`.
-- **`--format=json`** — projected row view. Same JSON structure but built from the CLI's column model (`--fields` applies). Use this when you only want specific fields.
-- **`--format=tsv|yaml|jsonl|id`** — tabular text formats; `--fields` filters columns.
+- **`--json`** — raw API payload passthrough. Errors are also JSON on **stdout**: `{ "schemaVersion": "1.2", "error": { "code", "kind", "message", "hint?" } }`.
+- **`--format=json`** — projected row view; `--fields` applies.
+- **`--format=tsv|yaml|jsonl|id`** — tabular text formats.
 
 ```bash
-# Raw API payload (--json)
 switchbot devices list --json | jq '.deviceList[] | {id: .deviceId, name: .deviceName}'
-
-# Projected rows with field filter (--format)
 switchbot devices list --format tsv --fields deviceId,deviceName,type,cloud
 switchbot devices list --format id      # one deviceId per line
-switchbot devices status <id> --format yaml
 ```
 
 ## Cache
 
-The CLI maintains two local disk caches under `~/.switchbot/`:
-
-- `devices.json`: Device metadata (id, name, type, category, hub, room…).
-  Default TTL: 1 hour.
-- `status.json`: Per-device status bodies.
-  Default TTL: off (0).
+Two local disk caches under `~/.switchbot/`:
 
-The device-list cache powers offline validation (command name checks, destructive-command guard) and the MCP server's `send_command` tool. It is refreshed automatically on every `devices list` call.
-
-### Cache control flags
+| Cache | Default TTL | Purpose |
+|---|---|---|
+| `devices.json` | 1 hour | device metadata; powers offline validation |
+| `status.json` | off | per-device status; GC'd after 24h |
 
 ```bash
-# Turn off all cache reads for one invocation
-switchbot devices list --no-cache
-
-# Set both list and status TTL to 5 minutes
-switchbot devices status <id> --cache 5m
-
-# Set TTLs independently
+switchbot devices list --no-cache          # bypass for one invocation
+switchbot devices status <id> --cache 5m   # set list + status TTL
 switchbot devices status <id> --cache-list 2h --cache-status 30s
-
-# Disable only the list cache (keep status cache at its current TTL)
-switchbot devices list --cache-list 0
 ```
 
-### Cache management commands
-
-```bash
-# Show paths, age, and entry counts
-switchbot cache show
-
-# Clear all cached data
-switchbot cache clear
-
-# Scope the clear to one store
-switchbot cache clear --key list
-switchbot cache clear --key status
-```
-
-### Status-cache GC
-
-`status.json` entries are automatically evicted after 24 hours (or 10× the configured status TTL, whichever is longer), so the file cannot grow without bound even when the status cache is left enabled long-term.
-
 ## Exit codes & error codes
 
 - `0`: Success (including `--dry-run` intercept when validation passes).
@@ -1209,112 +807,22 @@ npm install
 
 npm run dev -- <args>       # Run from TypeScript sources via tsx
 npm run build               # Compile to dist/
-npm test                    # Run the Vitest suite (2204 tests)
+npm test                    # Run the Vitest suite (2225 tests)
 npm run test:watch          # Watch mode
 npm run test:coverage       # Coverage report (v8, HTML + text)
 ```
 
-### Project layout
-
-```text
-src/
-├── index.ts              # Commander entry; mounts all subcommands; global flags
-├── auth.ts               # HMAC-SHA256 signature (token + t + nonce → sign)
-├── config.ts             # Credential load/save; env > keychain > file priority
-├── api/client.ts         # axios instance + request/response interceptors;
-│                         # --verbose / --dry-run / --timeout wiring
-├── credentials/
-│   ├── keychain.ts       # Credential store interface + OS backend selection
-│   └── backends/         # macos.ts / linux.ts / windows.ts / file.ts
-├── devices/
-│   ├── catalog.ts        # Static device catalog (commands, params, status fields)
-│   └── cache.ts          # Disk + in-memory cache for device list and status
-├── install/
-│   ├── steps.ts          # Generic step runner with rollback support
-│   ├── preflight.ts      # Pre-flight checks (Node, npm, network, agent)
-│   └── default-steps.ts  # Concrete steps: credentials, keychain, policy, skill, doctor
-├── policy/
-│   ├── validate.ts       # Schema version dispatch + JSON Schema validation
-│   ├── migrate.ts        # v0.1 → v0.2 migration
-│   ├── load.ts           # YAML file loading + error handling
-│   ├── add-rule.ts       # Rule injection into automation.rules[]
-│   ├── diff.ts           # Structural + line diff
-│   └── schema/v0.2.json  # Authoritative v0.2 JSON Schema
-├── rules/
-│   ├── engine.ts         # Main orchestrator (MQTT + cron + webhook)
-│   ├── matcher.ts        # Trigger + condition matchers
-│   ├── action.ts         # Command renderer + executor
-│   ├── throttle.ts       # Per-rule throttle gate
-│   ├── cron-scheduler.ts # 5-field cron + days filter
-│   ├── webhook-listener.ts # HTTP listener (bearer token, localhost-only)
-│   ├── pid-file.ts       # Hot-reload via SIGHUP or sentinel file
-│   ├── audit-query.ts    # Audit log filtering + aggregation
-│   ├── conflict-analyzer.ts # Static conflict detection (opposing actions,
-│   │                     #   high-freq MQTT, destructive cmds, quiet-hours gaps)
-│   ├── suggest.ts        # Heuristic + LLM-backed rule YAML generation
-│   ├── notify.ts         # notify action executor (webhook / file / openclaw)
-│   └── types.ts          # Shared rule/trigger/condition/action types (CommandAction | NotifyAction)
-├── llm/
-│   ├── index.ts          # createLLMProvider factory + LLM_AUTO_THRESHOLD
-│   ├── complexity.ts     # Intent complexity scorer (0–10) for auto-routing
-│   ├── rule-prompt.ts    # System prompt builder (embeds v0.2 schema snippet)
-│   └── providers/
-│       ├── openai.ts     # OpenAI-compatible provider (uses Node.js https)
-│       └── anthropic.ts  # Anthropic provider
-├── status-sync/
-│   └── manager.ts        # Spawn/stop logic, state file, OpenClaw bridge
-├── lib/
-│   └── devices.ts        # Shared logic: listDevices, describeDevice, isDestructiveCommand
-├── commands/
-│   ├── auth.ts           # `auth keychain` subcommand group
-│   ├── config.ts
-│   ├── devices.ts
-│   ├── expand.ts         # `devices expand` — semantic flag builder
-│   ├── explain.ts        # `devices explain` — one-shot device summary
-│   ├── device-meta.ts    # `devices meta` — local aliases / hide flags
-│   ├── install.ts        # `switchbot install` / `uninstall`
-│   ├── policy.ts         # `policy validate/new/migrate/diff/add-rule/backup/restore`
-│   ├── rules.ts          # `rules suggest/lint/list/explain/run/reload/tail/replay/
-│   │                     #   conflicts/doctor/summary/last-fired/webhook-*/
-│   │                     #   trace-explain/simulate`
-│   ├── scenes.ts
-│   ├── health.ts         # `health check/serve` — report + HTTP endpoints
-│   ├── upgrade-check.ts  # `upgrade-check` — npm registry version check
-│   ├── status-sync.ts    # `status-sync run/start/stop/status`
-│   ├── webhook.ts
-│   ├── watch.ts          # `devices watch <deviceId>`
-│   ├── events.ts         # `events tail` / `events mqtt-tail`
-│   ├── mcp.ts            # `mcp serve` (MCP stdio/HTTP server)
-│   ├── plan.ts           # `plan run/validate/suggest`
-│   ├── cache.ts          # `cache show/clear`
-│   ├── history.ts        # `history show/replay`
-│   ├── quota.ts          # `quota status/reset`
-│   ├── catalog.ts        # `catalog show/diff/path`
-│   ├── schema.ts         # `schema export`
-│   ├── doctor.ts         # `doctor`
-│   ├── capabilities.ts   # `capabilities`
-│   └── completion.ts     # `completion bash|zsh|fish|powershell`
-└── utils/
-    ├── flags.ts          # Global flag readers (isVerbose / isDryRun / getCacheMode / …)
-    ├── output.ts         # printTable / printKeyValue / printJson / handleError
-    ├── format.ts         # renderRows / filterFields / output-format dispatch
-    ├── audit.ts          # JSONL audit log writer
-    └── quota.ts          # Local daily-quota counter
-tests/                    # Vitest suite (2204 tests, mocked axios, no network)
-```
+Source layout: `src/commands/` (one file per command group), `src/devices/` (catalog + cache), `src/rules/` (engine, matcher, throttle, audit), `src/policy/` (validate, migrate, schema), `src/llm/` (providers), `src/utils/` (output, format, flags). Tests are in `tests/` and mirror the `src/` structure.
 
 ### Release flow
 
-Releases are cut on tag push and published to npm by GitHub Actions:
-
 ```bash
-npm version patch        # bump version + create git tag
+npm version patch        # bump + create git tag
 git push --follow-tags
+# then: GitHub → Releases → Draft → Publish
 ```
 
-Then on GitHub → **Releases → Draft a new release → select tag → Publish**. The `publish.yml` workflow runs tests, verifies the tag matches `package.json`, and publishes `@switchbot/openapi-cli` to npm with [provenance](https://docs.npmjs.com/generating-provenance-statements).
-
-See [`docs/release-pipeline.md`](./docs/release-pipeline.md) for the full pre-publish and post-publish verification flow (local hooks → CI → `publish.yml` → `npm-published-smoke.yml`).
+See [`docs/release-pipeline.md`](./docs/release-pipeline.md) for the full CI / publish verification flow.
 
 ## License