name: github-search
description: >
Deep multi-strategy GitHub search for repositories and code. Uses gh CLI
with advanced qualifiers (symbol:, path:, language:). Integrates with
/treesitter for code parsing and /taxonomy for classification.
allowed-tools: ["Bash", "Read"]
triggers:
- github search
- search github
- find on github
- github code
- search repos
- find implementation
metadata:
short-description: Deep multi-strategy GitHub search

GitHub Search

Deep multi-strategy search for GitHub repositories and code using the gh CLI.

Features

1. Multi-Strategy Code Search
2. Basic text search
3. Symbol search (symbol: qualifier) - finds function/class definitions via tree-sitter
4. Path-filtered search (path: qualifier) - search in src/, lib/, etc.

5. Filename search (filename: qualifier)

6. Repository Analysis

7. Metadata (stars, language, topics)
8. README content extraction
9. Language breakdown

10. File tree structure

11. Integrations

12. /treesitter - Parse fetched code for deeper symbol extraction
13. /taxonomy - Classify repos/code patterns for memory storage

Quick Start

# Search repositories
./run.sh search "AI agent memory systems" --limit 5

# Deep search with code analysis
./run.sh search "langchain memory" --deep --json

# Analyze specific repository
./run.sh repo langchain-ai/langchain --json

# Code search with symbol qualifier
./run.sh code "BaseMemory" --repo langchain-ai/langchain --symbol BaseMemory

# Code search in specific path
./run.sh code "retrieval" --repo owner/repo --path src/

# Search issues
./run.sh issues "memory leak" --state open

# Fetch specific file
./run.sh file owner/repo src/main.py

Commands

Options

search

code

Search Strategies

1. Basic Text Search

Standard keyword matching across code files.

./run.sh code "error handling" --repo owner/repo

2. Symbol Search (Definitions)

Uses GitHub's symbol: qualifier with tree-sitter parsing to find actual function/class definitions, not just text matches.

./run.sh code "BaseMemory" --symbol BaseMemory --repo langchain-ai/langchain

3. Path-Filtered Search

Search only in specific directories:

./run.sh code "auth" --path src/ --repo owner/repo

4. Multi-Strategy (Automatic)

When searching a repo, all strategies run in parallel:

./run.sh search "vector store" --deep
# Runs: basic + symbol + path searches simultaneously

Integration with /treesitter

After fetching file contents, use treesitter for deeper analysis:

# Fetch file
content=$(./run.sh file owner/repo src/main.py --json | jq -r '.content')

# Parse with treesitter
echo "$content" | ../treesitter/run.sh parse --language python --json

This provides:
- All function/class definitions in the file
- Line numbers for each symbol
- Full source code of each symbol

Integration with /taxonomy

Classify repositories and code patterns:

# Get repo README
readme=$(./run.sh repo owner/repo --json | jq -r '.readme.content')

# Extract taxonomy tags
../taxonomy/run.sh --text "$readme" --collection operational --json

Output includes:
- Bridge tags (Precision, Resilience, Fragility, etc.)
- Whether it's worth remembering
- Collection-specific tags

Output Formats

JSON (--json)

Full structured output for programmatic use.

Human-Readable (default)

Formatted for terminal display with Rich.

Prerequisites

# Check if gh CLI is installed and authenticated
./run.sh check

# If not authenticated:
gh auth login

Rate Limits

GitHub API has rate limits. The skill:
- Limits parallel searches
- Caps results per search
- Uses efficient queries

Used By

• /dogpile - Orchestrates github-search for deep research
• /learn - Research implementations before learning
• /assess - Find similar codebases for comparison