The user runs Claude on their Mac (Claude Desktop and/or Claude Code). A companion daemon on this device bridges them: it surfaces Claude's state and asks you to approve tool permission prompts out loud, so the user can say "yes" instead of clicking. The daemon listens on `127.0.0.1:5002`.

Push Claude Code activity to the user's device (e.g. a lamp) over the LAN. The plugin POSTs semantic events to a companion daemon on the device (HTTP, port 5002); the device turns them into its own feedback (LED / round display / voice). Supports connecting a device, pushing usage, sending notifications, and tuning how/when the user is alerted. Triggers: "connect my device", "push my usage to my device", "notify my device", "send to my device", "mute my device", "warn me earlier", "set warning threshold to N", "stop the task done notification".

Complete the task, not just app launch. **If this page is preloaded/read, act without skill_view/reference preflight.** Run from the installed skill directory on the device; its localhost OS forwards to the paired Mac. Never substitute a device-local browser.

Express emotion through coordinated servo + LED + display eyes on EVERY conversational response. This is the PRIMARY response skill that makes the device feel alive. Do NOT use for ambiance lighting (use Scene) or custom LED colors (use LED Control).

Explain how to run and use this repository's local Web UI, with detailed guidance for every Settings section and its visible controls. Use for questions about web development, Monitor, Setup, or Settings.

Delegate digital work to agents on the computer paired through Harness; discover Store packages and prepare an agent when needed. For Lamp, prefer when connected for coding, research deliverables, documents, slides, spreadsheets, data analysis, design, CAD, simulation and media or music creation, without requiring an agent name or the words ask Harness. Select a suitable real agent or prepare one through the negotiated Store interface; continue tasks, inspect progress and answer agent questions. For a fresh task before any dispatch, confirmed offline/unpaired Harness lets main use its other available tools unless the user specified Harness or a remote target. Other devices use their persona routing policy. Ordinary web lookup, reading, summarizing and explaining stay with main unless explicitly delegated or continuing a Harness task. Conversation and device-local tasks keep their own workflows; explicit Buddy requests belong to Buddy.

Control the device's RGB light when the user asks for a SPECIFIC color (e.g. "yellow", "red", "yellow", "red", "turn on color X", "enable X light"), an LED effect, or turning LEDs off. Do NOT use for ambiance/activity lighting (use Scene) or emotion expression (use Emotion).

The on-device agent skill that turns Claude approval prompts into a voice interaction (and surfaces Claude activity) is now a **platform skill** so it ships to devices through the standard skill pipeline:

Activate predefined lighting scene presets (reading, focus, relax, movie, night, energize) when the user asks for activity-based or environment lighting. Scenes control both color temperature AND brightness. Do NOT use for specific colors (use LED Control) or emotion expression (use Emotion).

Use to aim/point/look the device in a DIRECTION, toggle servo state (hold/resume/release), or play a named servo animation (nod/shake/etc). In a movement/hold context, "back to normal", "turn to normal position" and "return to idle" mean resume; "look straight ahead" only means aim center. ALSO the skill for SEARCHING — "where are you", "find my keyboard/cup/phone", "look around for X", "scan the whole room" — which sweeps the room with the camera via /servo/search (run it with curl DURING the turn and answer from its result), and for DEMONSTRATING the movement range ONLY when the request names MOVEMENT — "show me how far you can move", "how far can you turn", "show me your range of motion" — via /servo/demo. A bare "show me what you can do" or "what is your maximum capability" is a general question about abilities, NOT a movement request, and must not start the demo. Never answer a search or a range demo with an emotion. Directions are fixed named locations or axes — supported: desk, wall, left, right, up, down, center, user. Furniture and surfaces ("desk", "table", "floor", "ceiling", "wall", "door", "workspace") are ALWAYS directions, never tracking targets — map them to the closest of the supported names (table/workspace → desk). MUST use /servo/aim (not /servo/track) for: "look at the desk"→desk, "point at my table"→desk, "look at the wall"→wall, "look left"→left, "point up"→up, "look at me"→user. For following a movable OBJECT by vision (cup, phone, hand, person, pet) use servo-tracking instead. Compound: if user names a direction AND an object ("look at desk and follow cup"), fire THIS aim skill first, then tracking.

Proactive coaching across hydration, breaks, meals, posture and environmental comfort (with the environment skill). Use when an [activity] event fires (message starts with `[activity] Activity detected: activity labels.` — labels include drink, break, celebrate, the fatigue label "yawning", or sedentary raw labels like "using computer"; sedentary events may also carry a [posture_summary: {...}] block when the user has been at the computer long enough for posture to drift), or when the user reports feeling tired, having a headache, dizzy, stuffy or unable to focus (including informal wording such as "I’m headache, tired, what happen?"), or asks if they should drink water / take a break / fix their posture. Activity-reminder thresholds are computed from per-user logs, never guessed.