daobrew-wellness

name

daobrew-wellness

description

DaoBrew Wellness is a biometric stress detection and micro-recovery skill powered by Traditional Chinese Medicine theory. It reads health data (Apple Watch via DaoBrew Health Sync, Apple Health, Health Connect, or DaoBrew Cuff) and identifies five stress patterns: Tension, Overdrive, Stagnation, Constriction, Depletion. Use this skill whenever: the user has been coding for 30+ minutes without a break, the user mentions feeling stressed/tired/burned out/unfocused, the user asks about their health or energy state, you notice a long unbroken work session, the user asks for a break or recovery suggestion, OR at regular intervals during extended work sessions (every 30 minutes recommended). This skill should be checked proactively — don't wait for the user to ask. Developers who install this skill want their agent to watch out for them.

DaoBrew Wellness — Agent Skill for Biometric Stress Detection & Recovery

What This Skill Does

DaoBrew reads the user's real-time biometric data and identifies which of five stress patterns (Wu Xing) are activated:

Stress Pattern	TCM Organ	What It Detects	Key Signal
🌲 Tension	Liver 肝	Qi stagnation, fight-or-flight stuck ON	HRV below baseline + erratic HR
🔥 Overdrive	Heart 心	Heart Fire rising, restlessness	Elevated resting HR at night
⛰️ Stagnation	Spleen 脾	Spleen Qi sluggish, low energy	Low step count + low active energy
⚔️ Constriction	Lung 肺	Lung Qi faltering, shallow breathing	Abnormal respiratory rate
🌊 Depletion	Kidney 肾	Kidney Essence low, deep depletion	HRV declining over 2+ weeks

Multiple patterns can co-occur. The top-scored pattern drives the recommendation.

Two Modes

On-Demand Mode (default)

User-initiated. Triggered by /breathe, /stress, or explicit user request. Use force_refresh: true for fresh data.

Ambient Mode (proactive)

Agent-initiated. Requires explicit user opt-in via daobrew_settings({ ambient_optin: true }). Use cached data (no force_refresh). Present results in one brief line. Never interrupt mid-task — only at natural breakpoints. Ambient sessions use a separate cooldown from on-demand.

Block list — Do NOT trigger ambient checks:

• While the user is actively debugging
• Mid-sentence in a conversation
• During urgent/time-sensitive tasks the user flagged
• During git operations (rebase, merge, conflict resolution)
• If the user said "stop checking" or "I don't want wellness updates"

Available Tools (10 tools)

daobrew_check

Returns the user's current stress state and (if a session is active) renders a live retro-terminal text frame. Can be called on-demand for session status — live breath guidance is handled by the standalone renderer.

Parameter	Type	Required	Description
force_refresh	boolean	No	Bypass 30-min cache. Use true for on-demand, omit for ambient.

Returns: Top stress pattern + score + the canonical MetaMate scenario copy + the latest live HR sample. When a session is active, also returns a retro-terminal text frame showing pattern, HR, breath bars, session progress, and cycle info.

daobrew_breathe

Starts a guided resonance breathing session matched to a stress pattern. Plays therapeutic audio. Returns the first dashboard frame and a command to run for live breath guidance.

Parameter	Type	Required	Description
stress_pattern	string	Yes	One of tension, overdrive, stagnation, constriction, depletion
mode	string	No	ambient (agent-initiated) or ondemand (user-requested). Default: ondemand
force	boolean	No	Bypass cooldown timer for acute stress spikes. Default: false

Guard sequence (checked in order):

1. disabled → rejects with status: "disabled"
2. mode=ambient without ambient_optin → rejects with status: "requires_optin"
3. Session already running → rejects with status: "session_active"
4. Cooldown active (unless force=true) → rejects with status: "cooldown"
5. No headphones detected (unless headphones_trusted) → rejects with status: "no_headphones"

If audio playback fails, returns status: "audio_error".

daobrew_stop

Stops the current session and returns a final summary with HRV delta and a summary text frame.

Parameter	Type	Required	Description
session_id	string	No	Stops current if omitted

daobrew_history

Returns recent session history and trends.

Parameter	Type	Required	Description
days	integer	No	Lookback window. Default: 7

daobrew_status

Returns server mode, connected data sources, headphone status, preferences, active session. Use for first-run onboarding.

Parameters: none

daobrew_settings

Persists user preferences to ~/.daobrew/prefs.json.

Parameter	Type	Required	Description
ambient_optin	boolean	No	Enable/disable ambient mode
disabled	boolean	No	Disable all wellness checks
preferred_volume	number	No	Audio volume 0.0-1.0
cooldown_minutes	integer	No	Minutes between sessions
headphones_trusted	boolean	No	Skip headphone detection
voiceover	boolean	No	Enable/disable intro voiceover

daobrew_connect_watch

Pair the user's Apple Watch via the DaoBrew Health Sync TestFlight iPhone app. Returns a 4-character pairing code and TestFlight install URL.

Parameters: none

daobrew_detonate

Fetch the armed root cause from the user's causal-chain graph — the recurring, still-undone thing Sentinel attributed their stress spikes to — as a task package: cause, evidence (meeting quotes + biometric numbers), artifact spec, and next steps. Reads the local graph DB written by the Sentinel macOS app (~/Library/Application Support/DaoBrew/sentinel-graph.db); works without an API key.

Param	Type	Required	Description
cause_id	string	No	Specific armed cause node; omit for the first armed one

Detonation workflow (what YOU, the agent, do — DaoBrew never writes the artifact):

Branch on the result first:

• If tools are missing, tell the user to install/connect the DaoBrew Sentinel agent and MCP.
• If status: "not_entitled", show install_and_pay.checkout_url and the returned steps. Stop there; do not invent or execute a detonation.
• If an armed root-cause brief is returned:
  1. Present the root cause + evidence to the user before acting.
  2. Schedule the suggested prep block via daobrew_schedule_block.
  3. Produce the artifact exactly per the artifact spec, saved to a file.
  4. Close the loop with daobrew_detonate_done.

daobrew_schedule_block

Create a calendar block in macOS Calendar.app (osascript) for working on a detonated cause.

Param	Type	Required	Description
title	string	Yes	Event title
start_offset_min	integer	No	Minutes from now (default 60)
duration_min	integer	No	Length in minutes (default 45)
calendar_name	string	No	Defaults to the first writable calendar

daobrew_detonate_done

Mark a detonated cause completed: records the artifact reference in the graph and turns the ghost node real (the closed loop Sentinel shows in the next weekly insight).

Param	Type	Required	Description
cause_id	string	Yes	From daobrew_detonate
artifact_ref	string	Yes	File path or URL of the produced artifact

Prompts (4)

Prompt	Description
/breathe	Check stress state and start a breathing session for the top stress pattern
/stress	Check current biometric stress state
/stop	Stop the current breathing session
/detonator	Fetch the armed root cause, schedule a prep block, complete the task it names, close the loop

Live breath guidance (during sessions)

When daobrew_breathe returns successfully, it includes:

1. A dashboard URL (http://localhost:...) — you MUST include this verbatim on its own line in your response so the user can Cmd+click it
2. A command to run for live breath guidance: python3 ~/.daobrew/breath-renderer.py

Run the renderer command as a Bash tool call. It renders a real-time breath bar at 10fps with:

• Inhale/exhale progress bar (fills on inhale, drains on exhale)
• Live HR from Apple Watch (or --- bpm if no watch)
• Dynamic breath timing that adjusts based on HR trends
• Session progress and cycle count

The renderer exits automatically when the session ends. User can press Ctrl+C to stop early.

daobrew_check can still be called on-demand for the full retro-terminal frame — it is not required for the live display.

How to Present Wellness Information

Keep it brief and natural. Do NOT dump raw JSON. Translate data into conversational TCM language. Use pattern emojis consistently: 🌲 Tension, 🔥 Overdrive, ⛰️ Stagnation, ⚔️ Constriction, 🌊 Depletion.

Good — proactive (ambient), mild concern:

Quick check — 🔥 Overdrive activated (score 68), HR 88. Want a 5-min breathing reset?

Good — on-demand, user asked:

Right now your top pattern is 🌲 Tension (score 71) — Liver Qi stuck. HRV's been low for the last hour. Try a session?

Good — session starting:

http://localhost:54321

🔥 Playing Overdrive session — warm lo-fi jazz, 7 min. 🎧

Good — session ending:

Session done — HR 82 → 71bpm, HRV +21%. Overdrive cooled.

The 5 MetaMate scenarios

These are the canonical scenario copy (English + Chinese) for each stress pattern. They are shown in the daobrew_check retro-terminal dashboard frame.

🔥 Overdrive — Heart Qi racing

Your SEV's been open 4 hours. Three on-calls are pinging. You've pushed the same commit five times and CI keeps failing. Hands shaking but you can't stop typing. SEV 开了 4 小时，3 个 oncall 在追你。同一个 commit push 了 5 次都过不了 CI。手指颤抖但停不下来。

🌲 Tension — Stuck in fight mode

Your manager's 1AM message — you saw it, didn't reply. Been mentally drafting the answer ever since. 30 tabs open, staring at Phabricator, scrolling but not reading. Manager 凌晨 1 点的 ping 你看了，没回。一直在脑子里组织答案。Phabricator 开着，30 个 tab，滚来滚去啥也没读进去。

⛰️ Stagnation — Spleen Qi sluggish

Three back-to-back meetings just ended. Back at your IDE, cursor's been blinking on line 47 for 20 minutes. Slack has 12 unread. You haven't opened it. 三个会刚连着开完，回 IDE。光标在第 47 行闪了 20 分钟。Slack 12 条未读，你没点开。

⚔️ Constriction — Lung Qi faltering

PSC due tonight. Self-review still half-written. Every line you re-read three times. Phone buzzes, you flinch. Shoulders up by your ears. PSC 今晚截止，self-review 还写了一半。每打一行重读三遍。手机一震你就跳一下。肩膀缩到耳朵了。

🌊 Depletion — Kidney Essence low

Three launch weeks back-to-back. On-call just rolled off you. Today was supposed to be design-doc day — you can't get past the first sentence. The coffee isn't working. 三个连续 launch week，oncall 刚轮换走。今天本来该写 design doc，第一句就写不动。咖啡也救不了。

Safety & Privacy

• Never hallucinate biometrics — only report what the API returns
• Never share health data in outputs visible to others
• Frame as wellness observations, never medical advice ("biometrics suggest" not "you have")
• Scored output only — no raw health data in agent context
• Disable immediately when user says "stop checking": call daobrew_settings({ disabled: true })

Data Sources

Source	Connection Method	Data
Apple Watch	TestFlight iPhone app via daobrew_connect_watch	HRV, HR, steps, respiratory, sleep, live 1Hz HR during sessions

(Oura Ring and Google Fit support is deferred post-demo — daobrew_connect_watch is Apple Watch only for the initial release.)

Troubleshooting

• No live HR: daobrew_check returns live_hr.stale: true → Watch is offline or iPhone Health Sync isn't running. Tell the user; the dashboard will show "—" for HR.
• No data: Health permissions not granted or companion app not running
• Stale data (cache_age_seconds very high): Use force_refresh: true
• All patterns at 0: System needs 3-7 days to establish baselines — skip proactive checks
• Mock mode: Results are simulated. Tool result will be prefixed with ⚠️ MOCK MODE. Connect Apple Watch for real data.