Skip to content

feat: add optional live skill evolution smoke#76

Draft
ohkyuetaek wants to merge 2 commits into
NousResearch:mainfrom
ohkyuetaek:feat/phase1-live-smoke
Draft

feat: add optional live skill evolution smoke#76
ohkyuetaek wants to merge 2 commits into
NousResearch:mainfrom
ohkyuetaek:feat/phase1-live-smoke

Conversation

@ohkyuetaek
Copy link
Copy Markdown

Summary

  • Add an optional live provider smoke runner for Phase 1 skill evolution
  • Add a fixture Hermes skill repo and small golden dataset for the smoke path
  • Document API-free regression tests vs manual live-provider smoke runs
  • Ignore generated output artifacts

Dependency

Verification

  • .venv/bin/python scripts/run_live_skill_evolution_smoke.py --print-command-only
  • .venv/bin/python -m pytest tests/test_live_smoke_script.py -q → 5 passed
  • .venv/bin/python -m pytest -q → 150 passed, 11 DSPy deprecation warnings
  • git diff --check
  • Static added-line scan: no hardcoded secrets, shell=True/os.system, eval/exec, or pickle.loads
  • Independent pre-commit review: passed; only follow-up suggestions for future cost-control/usability hardening

Notes

  • The live smoke path is manual-only and may incur provider cost. It prints the underlying command safely with shell quoting and checks known provider credential env vars before live execution.

- Store skill text in DSPy signature instructions so GEPA can evolve it
- Use DSPy 3.2 GEPA constructor with reflection LM and max_full_evals
- Validate full SKILL.md artifacts and honor configurable output_dir
- Add API-free golden dataset regression coverage for emitted artifacts
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant