Skip to content

Skill Catalog Builder

Builds a bounded, collision-checked index of agent skills across discovery scopes before loading them

Stars
76,331
Licence
MIT
Files
4 files

What it does

Lets an AI agent host that loads skills from several folders (for example project and user folders) scan them in a fixed precedence order, read only the metadata at the top of each SKILL.md, and report duplicate skill names and skipped entries before any full instructions are loaded. It is a discovery and bookkeeping step that keeps the agent's context small and its choices predictable.

Great for

  • Developers building or configuring an agent host that reads SKILL.md folders from multiple directories
  • Teams who want deterministic, explainable choice when two skills share a name
  • Authors studying how progressive disclosure of skills works

Skip it if

  • Ordinary end users who just want to install and use a skill; the catalog builder is infrastructure for hosts, not a task skill
  • Hosts that only load a single skill folder, where collision handling adds nothing
  • Environments where the scripts cannot be reviewed first, since the bundled script runs on the local filesystem

Try asking

  • “Build a skill catalog from my project and user skill folders, highest precedence first, and report any name collisions”
  • “Scan these two skill directories and list which skills were omitted and why”
  • “Show me the catalog budget and the selected skills before activating any of them”
Step-by-step guide: how to use Skill Catalog BuilderInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 4 original files, the MIT licence and an attribution note.

    skill-catalog-builder.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip skill-catalog-builder.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip skill-catalog-builder.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Review assets/scope-policy.json and set your own scope order rather than trusting the example order
  • Read the JSON collisions and omitted arrays before activating any skill, as the SKILL.md instructs
  • Review scripts/build_catalog.py before running it, since it executes locally on your files

What's inside

  • assets/scope-policy.json
  • references/discovery-contract.md
  • scripts/build_catalog.py
  • SKILL.md

Redistributed unmodified from rohitg00/ai-engineering-from-scratch under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.