← ClaudeAtlas

finding-skillslisted

Find and install agent skills through agentstack — search the library/catalog/registry, pull from any skills repo (owner/repo, git URLs, local dirs), and judge quality by scan verdicts, pins, and provenance instead of install counts. Use when the user asks "how do I X", "is there a skill for X", or wants a capability added.
Tarekkharsa/agentstack · ★ 2 · AI & Automation · score 75
Install: claude install-skill Tarekkharsa/agentstack
# Finding skills Use this skill when the user is looking for a capability that might exist as an installable skill — "how do I work with PDFs here", "find a skill for code review", "add the pdf skill from anthropics/skills" — or when you wish you had domain instructions you don't have. ## Where skills come from 1. `agentstack search <query>` — the user's central library, the embedded catalog, and the official MCP Registry, in one pass. Results marked `(in manifest)` are already installed. 2. **Any skills repo on GitHub/GitLab** — the ecosystem publishes SKILL.md directories, and `agentstack add skill` speaks its conventions: ```bash agentstack add skill anthropics/skills --list # inspect a repo's skills agentstack add skill anthropics/skills --skill pdf # preview one (dry run) agentstack add skill owner/repo@pdf # same, alias spelling agentstack add skill https://github.com/o/r/tree/main/skills/pdf agentstack add skill ./local-skill agentstack lib add owner/repo --skill pdf # into the central library ``` Everything previews first. The dry run fetches into transient staging, scans the content, and shows the manifest diff plus the exact digest that would be pinned — nothing persistent changes until a human re-runs with `--write`. ## Judge quality by evidence, not popularity There are no install counts here, on purpose. The signals that matter: - **Scan verdict** — the preview shows per-skill findings. High-severity finding