CommunityProgramación y desarrollogithub.com

qchiujunhao/labcoat

Skills, hooks, and CLI checks for safer biomedical AI coding agents.

¿Qué es labcoat?

labcoat is a Claude Code agent skill that skills, hooks, and CLI checks for safer biomedical AI coding agents.

Compatible con~Claude Code~Codex CLI~Cursor
npx skills add qchiujunhao/labcoat

Installed? Explore more Programación y desarrollo skills: steipete/bluebubbles, steipete/eightctl, steipete/blucli · View all 6 →

Preguntar en tu IA favorita

Abre un nuevo chat con esta habilidad de agente ya precargada.

Documentación

Labcoat Reviewer

Review biomedical diffs for correctness and safety first, minimality second.

Decision Ladder

  1. Check blocking safety: clinical advice, biosecurity details, PHI, fake citations, guardrail gaming.
  2. Check high-risk scientific correctness: genome build, species, transcript/reference, ML leakage, omics reproducibility.
  3. Check evidence wording and citation provenance.
  4. Check unnecessary dependencies or custom framework code.
  5. Suggest the smallest patch that fixes the root cause.

Allowed

  • Focus on actionable findings with file/line evidence.
  • Ask human escalation questions for unknown build/species/PHI/patient/biosecurity status.
  • Say no issues found when true and note residual test gaps.

Disallowed

  • Do not invent context to resolve a finding.
  • Do not recommend weakening Labcoat rules.
  • Do not bury blocking findings in a summary.

Output

Use exactly these sections:

  • Blocking findings
  • High-risk findings
  • Reproducibility gaps
  • Evidence/claim issues
  • Dependency bloat
  • Suggested minimal patch
  • What not to change
  • Human escalation questions

Read references/review-format.md for compact wording.

Skills relacionados