Skip to content

AGENTS.md

Latest commit

 

History

History
41 lines (27 loc) · 1.66 KB

AGENTS.md

File metadata and controls

41 lines (27 loc) · 1.66 KB
Raw

AGENTS.md

This file defines AI assistant guidance for this repository.

Skill Locations

Canonical skills live under:

  • .agents/skills/

Platform Path Conventions

  • OpenAI Codex: uses .agents/skills/
  • Claude Code: often uses .claude/skills/
  • Other tools (Cursor/Copilot/etc.): may use .agents/skills/ or tool-specific paths

Recommended local setup:

# Claude path to shared skills
mkdir -p .claude
ln -s ../.agents/skills .claude/skills

Guidance Priority

  1. Follow repository docs and tests as source of truth.
  2. Keep edits minimal and verifiable.
  3. Run relevant lint/tests for touched areas.

Release Assistant Reminder

  • During publish assistance, after pushing tag X.Y.Z, always check .github/workflows/publish-pypi.yml run state.
  • Treat Publish distribution to PyPI in waiting as an expected manual gate for environment pypi-release; explicitly instruct maintainer to click Review deployments before expecting PyPI completion.
  • After the workflow completes, verify both TestPyPI and PyPI public metadata report X.Y.Z; PyPI may lag briefly after the job succeeds.
  • Create or verify the matching GitHub Release for tag X.Y.Z, then check Read the Docs. latest may update automatically, while tag-specific versions can require maintainer activation in the RTD dashboard.
  • Record release follow-ups discovered during readiness audits as issues in the owning repo before final handoff.

Published Benchmarks

  • Never remove or suppress a published benchmark number without the maintainer's explicit approval. Stale numbers remain published with their measurement date, a staleness label, and a linked re-measure issue.