biomedical-ai-skills: Instructions file for Claude Code

Biomedical AI Skills

Validated SKILL.md files for cancer biology and multi-omics research. Published to GitHub and to PyPI as biomedical-ai-skills.

This file is binding. Where it conflicts with habit, convenience, or a plausible-looking tutorial, this file wins.

Layout

skills/<name>/SKILL.md      the skill
skills/<name>/README.md     overview, diagram, gotchas table
skills/<name>/tests/        validation against public datasets
src/biomedical_ai_skills/   pip package and CLI
assets/                     images referenced by README, absolute URLs only

Not every directory under skills/ is published. Some are unreviewed local drafts excluded from git. Run git ls-files skills/ before claiming a skill exists publicly.

---

The five non-negotiables

1. Verify before you write. Never write a version number, function signature, argument name, or default value from memory. Check it against the registry and the package source, in this session, before it goes in a file.
2. Changing a skill means releasing. A skill edit that is committed but not released leaves pip install serving stale content. The release is part of the change, not a follow-up.
3. Never report success from an intermediate signal. A green workflow, a 200 response, or a successful commit is not evidence the artifact is correct. Verify the artifact itself.
4. Never claim a test passes unless you executed it. If it was not run, say "written but not executed" and say why.
5. Stage explicit paths. Never git add . or git add -A. Loose personal files sit in the repo root.

---

Verification gates

Each gate is a command. Run it and read the output. Do not proceed on assumption.

Gate 1 — before writing about any package

curl -s https://crandb.r-pkg.org/<pkg>        | python3 -m json.tool | head -20   # R
curl -s https://pypi.org/pypi/<pkg>/json      | python3 -m json.tool | head -20   # Python

Then read the actual signature from source, not a tutorial: