Thanks. Keep it tight.
- open or find the issue first
- keep one branch per issue
- read the workflow docs in
workflows/
Use these, not your imagination:
workflows/GITHUB-ISSUE.mdworkflows/GITHUB-PR.mdworkflows/GIT-COMMIT.mdworkflows/RELEASE.md
Run the obvious stuff before you ask anyone else to look:
./deva.sh --help
./deva.sh --version
./claude-yolo --help
./scripts/version-check.sh
shellcheck deva.sh agents/*.sh docker-entrypoint.sh install.sh scripts/*.shIf you change Docker image behavior, auth flows, or release logic, test those paths directly. Do not ship "should work".
- small, focused changes
- direct docs
- boring shell scripts that still work tomorrow
- explicit auth and mount behavior
- no surprise regressions
- prompt-engineering fluff in docs
- magical wrappers around simple shell code
- untested auth changes
- random formatting churn
- force-push chaos on shared branches
Update docs when behavior changes:
README.mdfor the front page and project positioningdocs/for long-form user guidesCHANGELOG.mdfor release notesDEV-LOGS.mdfor significant workSECURITY.mdwhen the reporting path or threat model changes
A good PR does three things:
- says what changed
- says why it changed
- says how you tested it
If it touches auth, container boundaries, or release mechanics, include the exact command you ran.
Do not freestyle releases.
Follow workflows/RELEASE.md. Update version, changelog, and docs together, then tag the release. If the tree is dirty and you do not understand why, stop.