Codex Chef Operator
Use this skill when maintaining the Codex Chef repository or a derived local setup.
Reference Routing
Read references/repo-maintenance.md before broad setup, installer, README,
catalog, docs, verification, release-readiness, or public-repo changes. Keep
SKILL.md as the routing entry point and put detailed repo policy in that
reference.
Workflow
- Inspect current files before changing anything.
- Preserve security defaults:
- sandbox stays enabled
- approvals stay interactive
- authenticated connectors stay disabled until needed
- no secrets, sessions, memories, auth files, or local project paths are added to source
- Keep docs and templates aligned:
AGENTS.mdREADME.mdREADME.tr.mdtemplates/codex/AGENTS.mdtemplates/codex/config.windows.tomltemplates/codex/config.unix.tomldocs/how-to.mddocs/how-to.tr.mddocs/skills-and-agents.mddocs/skills-and-agents.tr.mddocs/mcp-catalog.mddocs/mcp-catalog.tr.mddocs/codex-flags.mddocs/codex-flags.tr.mddocs/public-readiness.mddocs/public-readiness.tr.mddocs/install.mddocs/install.tr.mddocs/security-model.md- every English operator doc in
docs/should have a matching.tr.mdpair - the six root README entry points should route readers honestly to the complete English and Turkish operator docs
- bundled local skills should stay reference-backed, zero-network unless their own script or workflow explicitly says otherwise, and validated by repo checks
- public README visuals under
assets/should stay purposeful, accessible, lightly animated with reduced-motion fallback, and public-safe
- Validate after edits:
npm run validatenpm run validate:plugin-skillsgit status --short- Gitleaks when available
- Do not commit, push, tag, release, deploy, publish, rotate credentials, or perform account-level changes unless the user explicitly approves the exact action.
Output
Report:
- changed files
- validation commands and status
- remaining security or publishing risk