From ai-brain-starter
Fetches biometric context (HRV, sleep, recovery) from health-mcp DuckDB and injects it into journal, coaching, panel, patterns, and insights skill prompts. Skips silently when no data is available.
How this skill is triggered — by the user, by Claude, or both
Slash command
/ai-brain-starter:health-contextThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Wraps the vault-aware tools in `health-mcp` so other skills can pull biometric context without each one re-implementing the connection logic.
Wraps the vault-aware tools in health-mcp so other skills can pull biometric context without each one re-implementing the connection logic.
This skill is invoked WITH another skill, never standalone. It's the "look up what the body was doing during this moment" layer.
/journal or /daily-journal: pull health_journal_context(today) and prompt with HRV, sleep duration, sleep efficiency, steps, workouts/coaching: pull health_coaching_context(start, end, vault_root) for the coaching window/panel or strategy/decision question fires the advisory panel: pull health_panel_context(today, vault_root) for delta-vs-7d/patterns: pull health_floor_correlation(metric, days, vault_root) for HRV-vs-Floor and other biometric-vs-Floor patterns/weekly or /monthly (insights): pull health_weekly_rollup(week_start) and fold into the reviewDo NOT use for:
/ingest-health first)If health-mcp is not registered, has no data, or the tool call errors:
When the journal skill assembles its prompt, call health_journal_context(date) for today. Add to the prompt:
Body context (today, from health-mcp):
HRV: 28ms (vs 42ms 30-day baseline -- 33% below)
RHR: 65 bpm
Sleep: 5h 12m asleep (efficiency 87%, REM 38m, deep 22m)
Steps: 4,820
Workouts: 0
Mindful: 0 min
This becomes a journal-prompt input: "Your body had a tough night. How does that map to what you noticed?" The journal skill owns the prompt; health-context owns the data fetch.
When the coaching skill spans a multi-day window (e.g. last 14 days of accumulated tension), call health_coaching_context(start, end, vault_root). Surface:
This gives the coaching skill a "body track" alongside the emotional track.
When a panel fires on a strategy/decision moment, call health_panel_context(today, vault_root). Surface:
If delta < -10 AND today's floor_level is in the lower third, the panel can fold "your body and floor are both low; consider deferring high-stakes calls" into its synthesis. Never override the user's decision; only inform.
When the patterns skill runs, call health_floor_correlation(metric, days, vault_root) for each of: HRV, RHR, sleep duration, steps, recovery score. Surface any correlation with n >= 10 and |r| >= 0.25. Examples the patterns skill can render:
When the insights skill runs /weekly or /monthly, call health_weekly_rollup(week_start) for each week in the review. Add to the rendered review:
Every vault-aware tool needs vault_root. Resolve it from:
VAULT_ROOTNever hardcode a path. Pass as a tool argument every call.
This skill does not have its own / slash command. It piggybacks on the host skill via the host skill's invocation. The host skill MUST call health-context at the start of its prompt-assembly phase, before LLM synthesis, so the body data is available throughout.
Example (inside daily-journal's orchestrator):
# Inside daily-journal
try:
body = await call_tool("health_journal_context", {"date_str": today.isoformat()})
prompt += f"\nBody today: HRV {body['hrv_ms']}ms, RHR {body['rhr_bpm']} bpm, sleep {body['sleep_asleep_min']}min..."
except Exception:
# health-mcp not available; continue without
prompt += "\n(health context not available; run /ingest-health to set up)"
Reads only. Never writes the vault, never writes the health DB, never sends data over the network. Health Auto Export TCP-live is the one exception, and that's local-Wi-Fi only.
npx claudepluginhub mycelium-hq/ai-brain-starter --plugin ai-brain-starterShows freshness of health data sources (Apple Health, Oura, Fitbit, labs), last coach prescription and completion, missed-day count, lab status flags, and auto-trigger hook installation status for debugging the health-mcp auto-trigger chain.
Imports and standardizes health data from Apple Health, Fitbit, Oura Ring, CSV/JSON; integrates WellAlly.tech knowledge base for data management and personalized article recommendations.
Guides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.