majiayu000/claude-skill-registry

Searchable Claude Code skills catalog with source-linked guides and generated registry artifacts

Claude Skills Registry – What it is

A public, searchable catalog of Claude‑Code “skills” (re‑usable code snippets, tools, and agents that Claude can call). The project consists of three linked repositories:

  • claude-skill-registry-core – the authoritative pipeline that crawls skill sources, validates them, builds indexes and the static site.
  • claude-skill-registry-data – the raw archive of individual skill directories (skills/**).
  • claude-skill-registry – a generated mirror that bundles the core logic, the data, and the pre‑built JSON artifacts for downstream consumption.

The registry does not ship any skills itself; it only indexes and describes them, pointing back to the original repositories and licenses.

Why it matters

Claude‑Code agents need a way to discover community‑contributed capabilities (e.g., PDF parsing, test‑automation, DevOps helpers). This registry provides:

  • Fast, client‑side search – a gzipped JSON index that can be loaded in a browser or CLI.
  • Metadata & quality signals – star counts, security‑scan results, category tags, and install paths.
  • Multiple consumption methods – a web UI, a Go‑based CLI (sk), and a raw JSON API.

How to use it

Method What you get Quick start
Web UI Interactive search page (GitHub Pages). Open https://majiayu000.github.io/claude-skill-registry/
CLI (sk) Terminal‑based search and install of a single skill. bash\ngo install github.com/majiayu000/claude-skill-manager@latest\n# rename to `sk` and ensure it’s on $PATH\nsk search testing # find skills\nsk install anthropics/skills/skills/docx # install one\n
JSON API Direct download of the full index, category shards, quality/security rankings, etc. bash\ncurl https://majiayu000.github.io/claude-skill-registry/search-index-lite.json\n# or fetch the manifest and individual shard files\n

When you decide to install a skill, you follow the source link shown in the entry and respect its original license. The registry only points to the skill; it does not re‑license the content.

Architecture at a glance

Data Collection (core) → Index Generation (core) → Consumption (Web UI, sk CLI, JSON API)
  • Layer 1 – Data Collection – GitHub crawlers discover skill repos, download their files, and run a security scanner.
  • Layer 2 – Index Generation – Scripts produce several JSON artifacts:
    • search-index.json (full searchable index)
    • search-index-lite.json (compact version for fast client‑side use)
    • Category manifests, quality, security, and ranking indexes.
  • Layer 3 – Consumption – The generated files are served via GitHub Pages, the sk Go binary, or raw HTTP endpoints.

The core repo holds all pipeline code (scripts/, sources/, taxonomy/). The data repo stores the actual skills/** tree. The main (mirror) repo is a daily‑generated artifact that bundles the two for downstream users.

Getting started (for developers)

  1. Clone the core repo (the source of truth for the pipeline):
    git clone --filter=blob:none --sparse https://github.com/majiayu000/claude-skill-registry-core.git
cd claude-skill-registry-core
git sparse-checkout set --cone docs scripts sources schema
  2. Set up the Python environment (uses uv):
    uv venv && source .venv/bin/activate
uv pip install -r requirements.txt
  3. Build the search index locally (optional, to test changes):
    python scripts/build_search_index.py --registry registry.json --output docs
  4. Run the web UI locally:
    cd docs && python -m http.server 8000
# visit http://localhost:8000
  5. Submit new skills – either open an issue with the “Add Skill” template or PR a JSON entry to sources/community.json in the core repo.

Roadmap highlights

  • ✅ Daily automated updates, security scanning, and web UI are live.
  • ✅ Lightweight gzipped search index is available.
  • Planned: vector‑based semantic search, skill recommendation engine, version tracking, quality scoring, and API rate‑limiting.

License

The registry code and pipeline are released under the MIT License (see LICENSE).

Individual skills retain their original licenses; the registry does not re‑license third‑party content.

Related ecosystem

The registry is part of a broader “agent infra stack” for Claude‑Code and other LLM agents:

Layer Project Role
Extend claude-skill-registry Discover and browse community skills
Extend spellbook Cross‑runtime skill bundles
Trust argus Supply‑chain security scanner
Trust vibeguard Runtime guardrails for agents
Remember remem Persistent local memory
Orchestrate harness Rust‑based agent orchestration
Route litellm‑rs High‑performance LLM gateway
Keep keepline Session monitoring & recovery

TL;DR

The Claude Skills Registry is a maintained, searchable index of community‑contributed Claude‑Code plugins. It offers a web UI, a Go CLI (sk), and a raw JSON API, all built from an automated pipeline that crawls skill repositories, validates them, and publishes daily updates. The project is fully in scope for AI/LLM tooling and is open‑source under MIT.

Related

  • Project
    amitshekhariitbhu/ai-system-design

    The repository is a comprehensive, beginner‑friendly guide that teaches engineers how to design, build, and scale production AI systems around large language models, covering inference, hardware, caching, routing, RAG, agents, safety, observability, cost optimisation, and interview preparation.

  • Project
    Picovoice/cheetah

    Cheetah is Picovoice’s on‑device, streaming speech‑to‑text engine. It runs locally on phones, PCs, Raspberry Pi and browsers, supports several major languages, offers word‑level metadata, and provides SDKs for Python, C, iOS, Android, Flutter, React, Node, Java, .NET and Web. An AccessKey unlocks the library; you create an engine, feed PCM audio frames, receive partial transcripts, and flush for the final result. Demos and package builds are included for all platforms.

  • Project
    sveinbjornt/hear

    A command line interface for macOS that enables transcription of live microphone input and audio files using the system's built-in speech recognition.

  • Project
    typedef-ai/fenic

    fenic is a Python library that adds LLM‑driven semantic operators (extract, classify, summarize, embed, semantic join, etc.) to a familiar DataFrame/SQL‑style API. It lets you turn messy text, PDFs, logs, or JSON into typed rows, run the whole pipeline lazily with automatic batching, caching, and cost tracking, and then inspect or expose the result as a reusable tool for both humans and AI agents.