Issues and PRs welcome.
The most valuable contribution right now is testing the skill on platforms other than Claude. If you test on Gemini CLI, Codex, Copilot, Cursor, or any other agentskills.io-compatible platform, open an issue using the Compatibility Report template with:
- Platform and version
- Model used
- What worked
- What didn't
- Any workarounds applied
- Fork the repo and create a branch.
- Make your changes. If you modify files inside
skill-provenance/, follow the versioning protocol:- Update per-file
versionintegers inMANIFEST.yamlfor changed files. - Run
./skill-provenance/validate.sh --updateto recompute hashes. - Add a new top entry to
skill-provenance/CHANGELOG.mddescribing what changed. - Mirror the same release entry into root
CHANGELOG.md, which is the full archive. - Bump
bundle_versioninMANIFEST.yaml(PATCH for fixes, MINOR for features, MAJOR for breaking changes).
- Update per-file
- Run
./skill-provenance/validate.shto confirm everything is clean. - Open a PR with a clear description of what changed and why.
Changes to repo-level files that aren't part of the skill bundle (root README.md, AGENTS.md, CONTRIBUTING.md, .github/, LICENSE) don't require manifest updates or version bumps.
- SKILL.md and README.md: wrap prose at ~80 characters.
- YAML: 2-space indent.
- Bash:
set -euo pipefail, zero external dependencies. - JSON: 2-space indent.