davidliuk

davidliuk / graph-of-skills

Public

Dependency-Aware Structural Retrieval for Massive Agent Skills

claude-code knowledge-graph skills
91
11
100% credibility
Found Apr 13, 2026 at 54 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Graph of Skills builds a connected map of AI agent capabilities from skill documents to retrieve only the most relevant ones with their dependencies for any task.

How It Works

1
📖 Discover Graph of Skills

You hear about this helpful tool that organizes skills for AI assistants so they only get the most useful ones for each task.

2
🛠️ Set up your workspace

Follow simple steps to prepare your computer, like copying a settings file and adding a connection for smart matching.

3
📚 Gather skill collections

Download ready-made folders of skills that your AI can learn from.

4
🌐 Build the skill map

Create a connected map of skills that understands which ones depend on each other, making retrieval super smart.

5
🔍 Ask for task skills

Describe a task in plain words and instantly get a short list of the best matching skills with their helpers.

6
Put it to work
⚡
Quick test

Try it right away on sample challenges to see skills in action.

📊
Full evaluation

Run bigger tests comparing with and without the skill map.

🎉 Smarter AI assistant

Your AI now solves complex tasks faster and more reliably using just the right skills.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 54 to 91 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is graph-of-skills?

Graph-of-Skills is a Python library for dependency-aware structural retrieval over massive agent skills libraries. You feed it directories of SKILL.md files describing modular agent capabilities, and it builds an offline graph capturing dependencies and co-occurrences. At runtime, it retrieves compact, ranked bundles of relevant skills plus prerequisites, keeping agent contexts lean instead of dumping thousands of irrelevant ones.

Why is it gaining traction?

It beats plain vector retrieval by reranking with graph propagation, surfacing skills that chain together for complex tasks like coding or embodied AI. The CLI (gos index, gos retrieve) lets you index libraries and query instantly, with prebuilt workspaces on Hugging Face and Dockerized benchmarks for ALFWorld and SkillsBench. Developers dig the agent-focused pipeline: seed matches, merge, graph-rerank, cap output.

Who should use this?

Agent engineers scaling libraries for multi-step coding agents in SkillsBench-style Docker tasks. Embodied AI researchers evaluating retrieval in ALFWorld household sims. Teams hitting context limits with vanilla RAG on thousands of skills.

Verdict

Worth prototyping for dependency-aware agent retrieval—thorough docs, evals, and arXiv paper make setup painless despite 47 stars and 1.0% credibility signaling early alpha. Test on your skills corpus before production.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.