health-research

name: health-research description: Deep research on any health, biohacking, supplement, or medical topic. Triggers on "research", "what does science say", "evidence", "studies", "look into", or any request for scientific investigation of a health topic.

Health Research

Tiered research on any health topic with personalized context and proper filing.

Research Tiers

Automatic Tier Selection

When in doubt → Deep. More information is always better for health decisions.

Workflow

1. Load Context (aggressively!)

Always load ALL of these in parallel:

• docs/context/me.md — profile, baselines, health conditions
• docs/context/stack.md — current stack (interactions!)
• docs/context/watchlist.md — what we're tracking
• data/medical/README.md — lab tests, history
• domains/README.md — all domains + connections map

Also load ANY domain that might be related. When in doubt, load it.

2. Check Existing Research

BEFORE running any search, check if we already have research:

ls docs/research/

If relevant file exists:

• Read it
• Check if still valid (Query Context → Validity notes)
• If valid → USE IT, don't duplicate
• If outdated (weight changed, new labs, >3 months) → run new research referencing the old

3. Build Personalized Query

EVERY query (except Quick) MUST include user context:

PROFILE: [age, sex, height, weight, BF%, key conditions]
LABS ([date], at weight [X] kg): [relevant values]
CURRENT STATUS: [weight, BF%, relevant changes since labs]
STACK: [relevant items from stack.md]
HEALTH CONDITIONS: [active diagnoses, safety concerns]

Always ask about:

• Safety for user's specific conditions
• Interactions with current medications and supplements
• Evidence in relevant population (age, sex, condition)
• Dosing, timing, safety, contraindications

4. Run Research by Tier

Quick

• 1x perplexity_ask or perplexity_reason
• Answer directly, no file saved

Standard

• 1x perplexity_research (wide query covering the full topic)
• Save research file

Deep

Phase 1 — Wide: 1x perplexity_research — broad query covering the full topic

Phase 2 — Narrow (parallel): 3-6 targeted queries from this menu:

Phase 3 — Cross-Check: Compare narrow results with wide. Identify agreements, contradictions, new info.

Phase 4 — Save: Merge into single research file with cross-check section.

5. Save Research File

Save to: docs/research/[domain]/YYYY-MM-DD-topic.md

Required format:

# [Topic] Research

**Date:** YYYY-MM-DD
**Tier:** Quick / Standard / Deep
**Source:** Perplexity Research

---

## Query Context

**Input query (wide):**
> "[Exact query]"

**Narrow queries (if Deep):**
> 1. "[Query 1]"
> 2. "[Query 2]"

**Key context values:**
- Weight: X kg (from date Y)
- Lab results: from YYYY-MM-DD
- [Other relevant values]

**Validity notes:**
- If X changes, section Y may be outdated
- Valid until: [condition]

---

## Executive Summary
[3-5 sentences]

## [Sections by content]
...

## Cross-Check (if Deep)
| Claim | Wide | Narrow | Match? |
|-------|------|--------|--------|

## Actionable Items
[What specifically to do]

## References
[Preserve citations from Perplexity]

6. Update Domain + Watchlist

After saving:

1. Domain README — add key findings to Research section
2. Watchlist — if research recommends a lab test → add to watchlist
3. Connections — if cross-domain insight → check connections.md
4. Stack — if research recommends new supplement → note in stack.md if user agrees

Data Integrity

Never mix data from different time periods!

If using lab results:

1. Use body comp data from THAT period
2. Explicitly state: "Labs from [date] at weight [X] kg"
3. If things changed → note expected effect

If user provided an argument

/research magnesium → research magnesium (AI selects tier) /research magnesium deep → force Deep tier /research "is X safe?" → Quick tier /research (no argument) → ask for topic