Proteus is most useful when it captures real structural-biology workflow knowledge: command-line patterns, gotchas, reproducible helper scripts, and small tests that keep agents from repeating common mistakes.
- a PyMOL, ChimeraX, AlphaFold, RCSB, UniProt, or Rosetta gotcha with a tested fix
- a small stdlib-only helper workflow that returns JSON
- focused docs that help agents choose the right tool
- tiny fixtures and tests for edge cases
make test
python3 scripts/proteus_doctor.py --jsonKeep helper scripts dependency-free unless there is a clear reason not to.
Prefer deterministic JSON output in --json mode:
{"status": "ok", "data": {}}Errors should remain parseable:
{"status": "error", "error": "message"}- Keep changes scoped.
- Add or update tests for script behavior.
- Do not commit downloaded structures, private data, generated AlphaFold files, caches, or local machine artifacts.
- Document any external API or local binary assumptions.