Skills for OpenAI Codex CLI. Each skill constrains Codex to follow a proven workflow instead of improvising.
Copy any .md file from skills/ into your Codex skills directory:
cp skills/*.md ~/.codex/skills/codex-skillsOr copy one at a time:
cp skills/codex-quick-fix.md ~/.codex/skills/codex-skills| Skill | Purpose |
|---|---|
| quick-fix | Fix a single bug. Minimal change, run test, commit. |
| test-writer | Generate tests for a file. Detect framework, match patterns, cover edges. |
| refactor | Refactor with a safety net. Baseline tests first, then refactor, then verify. |
| api-scaffolder | Scaffold a REST API from a description or OpenAPI spec. |
| git-hygiene | Clean up a branch before PR: squash, rebase, fix messages. |
| dep-audit | Scan deps for vulns, outdated packages, unused imports. |
| migration | Upgrade a dependency to a new major version. |
| db-migration | Generate ORM migration files from a schema change. |
| ci-fixer | Read CI failure logs, fix the code, push, verify green. |
| self-review | Review changes before push: security scan, correctness check, test run. |
| prose-review | Edit prose to remove AI tells while keeping the writer's voice and every fact intact. |
Each file contains a prompt template and a checklist. When you invoke a skill, Codex follows the steps in order instead of guessing. The prompts use exact commands (grep patterns, test runners, git operations) so Codex can't skip steps.
The prose-review skill condenses three MIT-licensed skills that do this job in more depth. Install any of them straight into your Codex skills directory:
| Skill | Installs | Source |
|---|---|---|
| stop-slop | hardikpandya/stop-slop | |
| no-ai-slop | petergyang/no-ai-slop | |
| humanizer | blader/humanizer |
npx skills add hardikpandya/stop-slop
npx skills add https://github.com/petergyang/no-ai-slop --skill no-ai-slop
npx skills add blader/humanizerMIT
Vendored skill names in the table above link to their original MIT-licensed repositories and remain the work of their authors.