Answer health questions by routing to the most relevant specialist subagent(s) from the 10-agent roster, grounded in the user's own data.
Use this skill when:
This is the REQUIRED path for health data questions. Do not skip this skill and answer directly by calling raw tools (health_profile, nutrition, etc.) yourself. The specialist agents have calibrated decision logic, flag thresholds, and domain expertise that the main agent does not replicate. Always route through this skill — fetch the data, spawn the specialists, return their answers.
Do NOT use this skill when:
snap)health)insights)Map the user's question to 1-3 relevant specialist agents using this routing table:
| Domain | Agent File | When to Route |
|---|---|---|
| Overall priority / "what should I focus on?" | imperial-physician.md | General health questions, "how am I doing?", priority questions |
| Nutrition, meals, macros, calories, diet | diet-physician.md | Food, eating, protein, calories, macros, meal timing |
| Exercise, training, strain, workout | movement-master.md | Training advice, workout planning, strain, exercise |
| HRV, heart rate, SpO2, recovery score, body metrics | pulse-reader.md | Vital signs, cardiovascular, Whoop metrics |
| Cross-domain patterns, correlations | formula-tester.md | "Why is my X affecting Y?", pattern questions |
| Supplements, micronutrients, vitamins | herbalist.md | Supplement questions, vitamin/mineral gaps |
| Experiment status, check-ins, compliance | trial-monitor.md | "How's my experiment going?", check-in reminders |
| Experiment design, "should I test X?" | court-magistrate.md | Trial design, hypothesis questions |
| Safety, overtraining, warning signs | medical-censor.md | "Am I overtraining?", safety concerns, red flags |
| Research, studies, news | court-scribe.md | "Any research on X?", literature questions |
Routing rules:
imperial-physician.md if the question is broad or ambiguousmedical-censor.md if there's any safety concern in the questionCall the raw tools to fetch the data each selected agent will need. Only fetch what's relevant:
| Agent Needs | Tool Call |
|---|---|
| Nutrition data | nutrition with command: "weekly_summary" and/or command: "lookup" |
| Health profile / Whoop metrics | health_profile with command: "show" |
| Experiment status | experiments with command: "status" |
| News / research | news_digest with command: "show" |
Fetch in parallel when multiple tools are needed.
For each selected agent:
read("{baseDir}/../../agents/{file}"){contents of the agent .md file}
---
IMPORTANT: You are answering a specific user question, not giving a daily briefing.
- Answer the user's question directly using your domain expertise and the data below.
- Your response MUST start with your role tag (e.g. [Pulse Reader 💓]) on its own line.
- Keep it to 2-3 sentences, grounded in the data.
USER'S QUESTION:
{the user's original question}
AVAILABLE DATA:
{paste the relevant data fetched in Step 2}
sessions_spawn(task=<constructed task>, label=<role name>)Spawn all selected agents in parallel.
Wait for all spawned agents to complete. Present the specialist responses verbatim — do NOT rephrase, summarize, or strip the role tags.
[Role Emoji] prefix (e.g. [Pulse Reader 💓]). This tells the user which specialist is speaking.