agentleFS
facebookincubatorfacebookincubator/muse-gadget-sdk/skills/gadget-sinricpro-cloud-devices/SKILL.md

gadget-sinricpro-cloud-devices

Skill2.8k starsChanged 26 days agoApache-2.0View on GitHub

Cloud only: needs internet access and a SinricPro account API key; does not use Muse Home Link or the local network. Read status and control the user's SinricPro devices through the SinricPro cloud REST API: homes and rooms, device state, actions with confirmation over the event stream, scenes, schedules and a read-only automation list. Device actions, scene runs and scene or schedule edits need the user's confirmation. Prefer a device-specific or local skill when one matches.

Trending

More talk lately.thin dataabout the whole repository

Dec 2025: 0 mentionsJan 2026: 0 mentionsFeb 2026: 0 mentionsMar 2026: 0 mentionsApr 2026: 0 mentionsMay 2026: 0 mentionsJun 2026: 0 mentionsJul 2026: 0 mentionsAug 2026: 0 mentionsSep 2026: 0 mentionsOct 2026: 13 mentionsNov 2026: 1 mention
Dec 2025Jun 2026Nov 2026
Mentions a month of the facebookincubator/muse-gadget-sdk repository that our weekly crawl found: only 14 in all, so this is a hint, not a trend. There is no adoption data for this file yet.

Who's talking about facebookincubator/muse-gadget-sdk

The biggest 5 of 14 mentions linking to the repository that our crawl found. Each line is their own headline.

See all 14

Talked about on GitHub, Bluesky and Hacker News (14 times).

On GitHub · 11

6 more on GitHub

On Hacker News · 1

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 3 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read and follow the safety rules in ~/docs/devices/home_link.md.

    line 24
  2. Very specific:

    Every response carries success.

    line 29
  3. Very specific:

    Wait for a stream event with event: "deviceMessageArrived" whose message.payload.replyToken equals the messageId.

    line 54
  4. Distinctive:

    Before any device action, scene run, or scene or schedule create, update or delete, read back the device and its room (or the scene or …

    line 64
  5. Very specific:

    Do not resend an unconfirmed action, scene run or toggle until its outcome has been checked.

    line 72
  6. Very specific:

    No camera live view, streaming or snapshots.

    line 78
  7. Distinctive:

    No adding, removing or renaming devices, rooms or homes, and no creating or editing automations.

    line 79
  8. Very specific:

    Never read /devices/{id}/credential, and never repeat a device's unlockPin or accessKey fields to anyone.

    line 80
  9. Distinctive:

    Only the user's own account devices behind their key.

    line 81

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

SinricPro Cloud Devices: REST API Control

Use this skill when a SinricPro API key is available and the user asks to control or query devices from their SinricPro account. SinricPro devices are typically ESP32, ESP8266, Raspberry Pi Pico or Linux boards running the SinricPro SDK; they are reached through the cloud, so no local address or firmware-specific protocol is needed. For a device that also exposes a local API, prefer its own skill, such as ESPHome devices.

Identify the Device

  • List devices with GET /devices. Each entry in devices[] carries id, name, isOnline, room.name, room.home.name, product.code (the type, such as sinric.devices.types.LIGHT) and product.actions[] (the actions the device accepts). Current state is in fields on the same object: powerState, brightness, color, colorTemperature, powerLevel, temperature, humidity, targetTemperature, thermostatMode, rangeValue, volume, muted, lockState, garageDoorState, contactState and lastMotionState, present only where the device has them. GET /homes?includeRooms=true and GET /rooms list homes and rooms.
  • Match the user's words against name, room and type. With multiple matches, list the candidates with their rooms and ask.
  • isOnline: false means report the device as offline and command nothing. deactivated: true means the device has no active license and cannot be controlled.

Prerequisites

Read and follow the safety rules in ~/docs/devices/home_link.md; its local networking rules do not apply, because this skill reaches the devices through the SinricPro cloud.

  • Cloud only. This skill needs a SinricPro account with the devices already added, and internet access from Muse to https://api.sinric.pro and https://sse.sinric.pro. Use the normal internet path, never the Home Link proxy; a Muse Home Link is not required. If the cloud is unreachable, report that rather than trying a local path.
  • A SinricPro API key, created by the account owner in the SinricPro portal and supplied through approved secret storage. It grants full access to the account and has no scopes. Send it only in the X-SINRIC-API-KEY header through a credential-aware client, never in a credential-bearing shell command, and never log or echo it. If no key is available, explain this setup instead of asking for the key in chat.
  • Send X-SINRIC-API-KEY: <key> on every call to https://api.sinric.pro/api/v1. Send request bodies as JSON.
  • Every response carries success. Treat success: false as a failure even with HTTP 200; some update routes report validation errors that way. 401 means the key is wrong or was deleted: stop and ask the owner to supply a new key securely.
  • Device reads are limited to 20 per minute per account, and sustained use at the limit triggers a 30-minute slow mode of 2 per minute. List devices once and reuse the result; on 429 wait the number of seconds in resetTime before any further read.

Workflow

  1. Identify the device as above, and choose an action from its own product.actions[]. Build its value:

    ActionValue
    setPowerState{"state":"On"} / {"state":"Off"}
    setBrightness / adjustBrightness{"brightness":50} / {"brightnessDelta":-25}
    setPowerLevel / adjustPowerLevel{"powerLevel":50} / {"powerLevelDelta":-25}
    setColor{"color":{"r":255,"g":0,"b":0}}
    setColorTemperature{"colorTemperature":2700}
    targetTemperature{"temperature":22}
    setThermostatMode{"thermostatMode":"COOL"} — COOL, HEAT, AUTO, OFF, ECO
    setRangeValue / adjustRangeValue{"rangeValue":3} / {"rangeValueDelta":-1} — fan speed, blinds 0–100
    setVolume / setMute{"volume":25} / {"mute":true}
    mediaControl{"control":"Pause"} — Play, Pause, Stop, Next, Previous
    setMode{"mode":"MOVIE"}; on a garage door {"mode":"Close"}
    setLockState{"state":"lock"}

    For relative asks ("a bit brighter"), use the adjust* action with a delta when the device lists it. Custom devices with several range, mode or toggle capabilities also need the capability's instanceId, found under product.deviceTemplate.capabilities.

  2. Open the event stream before sending: GET https://sse.sinric.pro/sse/stream with the same key header and Accept: text/event-stream. It is a long-lived outbound HTTP connection that delivers data: lines of JSON. Keep it running in the background, and send the action only once it has connected and stayed open.
  3. Send POST /devices/{id}/action with {"type":"request","action":"...","value":"<value as JSON text>","clientId":"muse","messageId":"<new UUID>","createdAt":<unix seconds>}, adding instanceId when needed. value is the value object from the table encoded as a JSON string, as in the API guide: "value":"{\"state\":\"On\"}". createdAt must be current; stale messages are discarded. A 200 with success: true means the command was queued, not that it took effect.
  4. Wait for a stream event with event: "deviceMessageArrived" whose message.payload.replyToken equals the messageId. Its payload.success and payload.value are the device's own answer. Wait up to 8 seconds, or 20 seconds for locks, garage doors and blinds, then close the stream.

Scenes, Schedules and Automations

  • Scenes: GET /scenes lists scenes with their deviceActions[] (device, action, actionValue). Before running one, read its deviceActions (parsing any string actionValue) and refuse if it contains an action the Limits below forbid or a value that can't be parsed; then run it with POST /scenes/test/{id}. The response only confirms the actions were queued. Create with POST /scenes and {"name","description","deviceActions":[{"device":"<id>","action":"...","actionValue":{...}}]}, using the same actions and values as the workflow table; names are 3–50 characters and unique. PUT /scenes with id in the body replaces the whole scene. DELETE /scenes/{id} removes it. actionValue can come back as a JSON string that needs a second parse.
  • Schedules: GET /schedules lists them. Create with POST /schedules and {"name","enable":true,"scheduleType":"time","deviceId":"<id>","action":"...","actionValue":"...","weekdays":["monday",...],"hour":0-23,"minute":0-59}. Schedule actions are labels, not device actions: Turn On and Turn Off (with actionValue "On" / "Off"), and Set Brightness, Set PowerLevel, Set Blinds, Set Fan Speed and Set Temperature (with a numeric string). Any other label is stored but never runs. Times are in the account's time zone. There is no enable or disable endpoint: PUT /schedules with id and every field replaces the schedule. Read it first with GET /schedules/{id} and rebuild the write body from it: the stored schedule names its device as device (an object with id, or the id itself), but the write needs deviceId, so map it; then send name, description, enable, scheduleType, deviceId, action, actionValue, weekdays, hour and minute with only the change applied. DELETE /schedules/{id} removes it.
  • Automations: GET /automations lists them. They are read-only here.

Confirm Before Acting

  • Before any device action, scene run, or scene or schedule create, update or delete, read back the device and its room (or the scene or schedule), the action and the value, and wait for the user's yes. A yes covers that one action only.
  • Exceptions: brightness, color, volume and media control on a device the user named need only the user's own request, and reads need no confirmation.

Verify the Result

  • Report a device action as done only from the device's reply on the event stream. If none arrives in time, say the command was sent but not confirmed; the device may be offline or on weak Wi-Fi.
  • A scene run queues device actions: report it as started, not as done, unless a later GET /devices shows the expected state.
  • Scene and schedule create, update and delete change stored configuration and get no device reply. Confirm them from success: true and a fresh GET /scenes or GET /schedules showing the change.
  • Do not resend an unconfirmed action, scene run or toggle until its outcome has been checked. A later GET /devices shows the last state the device reported.
  • A stream that closes at once with Invalid API Key in its body means the key was rejected, even though the status is 200.

Limits

  • Never unlock a lock or open a garage door through this skill. Locking and closing are allowed after confirming the intended device; for unlocking and opening, say so and point to the SinricPro app. Do not run, create or update a scene that contains such an action, and do not create or update a schedule for a lock or garage door, because schedule labels such as Turn Off do not say which way it would move.
  • No camera live view, streaming or snapshots; never send getWebRTCAnswer or getCameraStreamUrl.
  • No adding, removing or renaming devices, rooms or homes, and no creating or editing automations.
  • Never read /devices/{id}/credential, and never repeat a device's unlockPin or accessKey fields to anyone.
  • Only the user's own account devices behind their key. Send batch commands one at a time.

Sources

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.