Portable, project-agnostic skills for Claude Code and OpenClaw — small, reusable developer-workflow tools that work in any repo. Cloned once, symlinked into your agent's skills directory, and versioned here so they're available on every machine.
Each skill is a self-contained folder with a SKILL.md (the agent reads its
frontmatter description to decide when to load it) and optional
REFERENCE.md / EXAMPLES.md / scripts/.
| Skill | What it does |
|---|---|
codex-review-loop |
Drive a PR to convergence through the Codex AI reviewer — build → PR → @codex review → verify each finding against HEAD → fix the real ones with regression tests → re-trigger until clean → human reviews last. |
pr-first-workflow |
Default to a branch + PR for every change (code and docs); branch → commit → PR → review → merge → return to the default branch; direct-to-main only on an explicit OK. Pairs with codex-review-loop. |
safe-prod-db-write |
Run one-off writes/backfills against a production DB safely — pull the connection into an mktemp file, dry-run, get explicit human authorization, execute, verify counts/invariants, clean up. |
On any machine — clone once, then run the installer:
git clone https://github.com/Digitizers/agent-skills.git ~/Documents/GitHub/agent-skills
cd ~/Documents/GitHub/agent-skills && ./install.shinstall.sh symlinks every skill folder into ~/.claude/skills/ (Claude Code).
It's idempotent and safe — it replaces a stale symlink and skips any real
directory of the same name, so it never clobbers unrelated skills. Pass a
different target for other agents:
./install.sh ~/.agents/skills # OpenClaw, or any custom skills pathThe symlink pattern keeps this repo the single source of truth — edit a skill
here, and every machine's agent picks it up. On a new machine: git clone +
./install.sh and every skill is back.
Drop a new skill-name/SKILL.md folder in, keep SKILL.md under ~100 lines,
give the description a clear "Use when …" trigger, and commit. Split detail
into REFERENCE.md when it grows.
MIT.