skill-retrieval

Index and retrieve skills via ColBERT semantic search.

10|1|Updated Dec 24, 2025

Search Tags:#indexing #semantic search #retrieval #skill #ColBERT

Ready to run

npx skills add https://github.com/zenbase-ai/code-voyager --skill skill-retrieval

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: skill-retrieval
Source: https://github.com/zenbase-ai/code-voyager/tree/main/.claude/skills/skill-retrieval
Command: npx skills add https://github.com/zenbase-ai/code-voyager --skill skill-retrieval

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Provides fast discovery of Claude skills using ColBERT-based semantic search and an indexed catalog.

Core Features & Use Cases

Build and query a skill index
Retrieve top matches with descriptions and scores
Discover skills that match a given task or phrasing

Quick Start

Build the index with voyager skill index --verbose, then search with voyager skill find "resume where we left off".

Frequently Asked Questions about skill-retrieval

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I search for skills using semantic search?▼

Semantic search over skill libraries uses ColBERT embeddings to match your query intent against indexed skills. Build an index with `voyager skill index --verbose`, then retrieve top matches with `voyager skill find` followed by your task description or phrasing.

What is ColBERT and why use it for skill retrieval?▼

ColBERT is an embedding-based retrieval method that performs fast, accurate semantic matching. It indexes skills once, then retrieves relevant matches by comparing query embeddings against skill embeddings, enabling quick discovery across multiple skill sources.

Can I index skills from multiple sources?▼

Yes. The skill retrieval system indexes and retrieves skills from plugins, local mirrors, generated skills, and user-created skills in a single consolidated catalog, supporting discovery across heterogeneous skill libraries.

What happens if the LLM or RAG components become unavailable?▼

The system includes graceful degradation—retrieval continues to function even when LLM-based metadata extraction or RAG components are unavailable, ensuring skill discovery remains operational.

How do I get started with semantic skill search?▼

Start by building the index with `voyager skill index --verbose`. Once indexed, search using natural language queries like 'find a skill for [task]' or 'what skill handles [problem]' to retrieve relevant matches with descriptions and relevance scores.