# Teamwork Research

> Use when a broad or deep external investigation needs multiple source classes, claim-level evidence synthesis, or contradiction resolution; do not use for a narrow lookup or local code inspection.

- Skill: `jinplu/teamwork-research` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add jinplu/teamwork-research`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jinplu/teamwork-research/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: JinPLu (https://skillmd.com/u/jinplu)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/jinplu/teamwork-research

---


# Teamwork Research

Keep narrow lookups native. When Research is activated, investigate the complete
question rather than creating a lightweight ritual around a few searches.

## Method

1. Name the decision this investigation serves, then define intended use, scope,
   freshness cutoff, privacy boundary, and material claims.
2. Search the strongest primary and current sources first, then add independent
   source classes where they can change confidence.
3. Connect each material claim to direct support, counterevidence, inference,
   confidence, date, and nearby citation.
4. Investigate contradictions and explain rejected sources or unresolved gaps.
5. Stop when the evidence distinguishes the served direction, the scope is
   answered, or further retrieval is unlikely to change the conclusion; state
   which condition applies. Results feed the matching discussion or plan
   section; they do not rewrite the whole case. Supplied sources are seeds, not
   automatic proof or a search boundary. Split work only across independent
   evidence areas and keep one synthesis owner.

Maintain an early provisional conclusion as evidence develops, clearly marked
as provisional. The research synthesis must preserve the decision use, a
question tree, a claim-to-evidence table, contradictions and unknowns, coverage
and stop conditions, and notes for every source that actually participated in
the reasoning; source notes are embedded in the document rather than delegated
to a separate index.

Return the supported conclusion, contradictions, residual uncertainty, coverage
result, and the explicit stop basis.

## Persistence

When a listed checkpoint fires, write in the same response cycle. If separate
stable identities each cross a checkpoint, write each to its own path.

Cross-chat memory lives in one Markdown document from `references/research.md`
at `docs/teamwork/research/<slug>.md`. Same identity means the same
research question; reuse that path and name the document you read. A different
subject gets a new path.

Checkpoints: a material claim-to-evidence synthesis is first settled;
contradictions change confidence or the conclusion; or a stop condition is
reached.