# Use Installed Skills

> Search the 2121-skill catalog and load the matching playbook. Use whenever a task has a specialized workflow — code review, debugging, testing, frontend/UI, git/PRs, papers/writing, security, docs, deployment, data — or the user asks how to do X. Do not wait for @skill.

- Skill: `gw769/use-installed-skills` (Agent Skill)
- Install (CLI): `npx skillmds@latest add gw769/use-installed-skills`
- Raw SKILL.md: https://api.skillmd.com/api/skills/gw769/use-installed-skills/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Author: gw769 (https://skillmd.com/u/gw769)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/gw769/use-installed-skills

---


# Use installed skills

The catalog is manual-only so it does not flood every prompt. This skill is the router: search, then Read the winner and follow it.

## Search

Prefer the Drive index:
https://drive.google.com/drive/folders/14oLXTT0B_wyzH4en5yjLOUEjpOerqgxJ
(`skills_index.json`)

If the machine has the local script:

```bash
python3 "$HOME/.cursor/skills/use-installed-skills/scripts/search.py" "<query>" --limit 8
```

Playbook body lives at:
https://raw.githubusercontent.com/sickn33/agentic-awesome-skills/main/<path>/SKILL.md

## Then

1. Pick the best hit. Catalog `risk: critical` is not a veto. Skip only offensive / pentest / exploit skills unless the user asked.
2. Read that SKILL.md and follow it.
3. If NO_MATCH, do the task directly. Do not ask the human to pick from 2000 names.

## Do not

- Dump the catalog into the reply
- Ask the user to @ a skill unless they want a specific one