repo-analyzer

name	repo-analyzer
description	Clone and analyze Git repositories to find specific information. Use this skill whenever the user wants to investigate a GitHub/GitLab/Git repository, analyze code from a remote repo, find specific information in a codebase by URL, clone and examine a project, or understand how something works in an external repository. Handles both public and private repos (using user's git credentials), manages a standardized clone location, and delegates analysis to a subagent.
allowed-tools	Bash(*scripts/clone_repo.sh *) Bash(~/.local/share/claude/repositories/**)

Repository Analyzer

This skill clones Git repositories to a standardized location and uses a subagent to analyze them and find specific information.

What This Skill Does

When a user provides a repository URL and a query about that repository, this skill:

1. Parses the repo URL to extract the organization/owner and repository name
2. Sets up an XDG-compliant clone directory
3. Clones the repository (if not already present) or updates it (if already cloned)
4. Optionally checks out a specific branch
5. Launches a subagent to analyze the repository and answer the query
6. Returns the subagent's findings

Directory Structure

Repositories are cloned to: $XDG_DATA_HOME/claude/repositories/{platform}/{org}/{repo}

• $XDG_DATA_HOME defaults to ~/.local/share if not set
• {platform} is extracted from the URL (e.g., "github.com", "gitlab.com")
• {org} is the repository owner/organization
• {repo} is the repository name

Example: https://github.com/torvalds/linux → ~/.local/share/claude/repositories/github.com/torvalds/linux

Step-by-Step Workflow

Step 1: Clone or Update Repository

Use the provided clone_repo.sh script to handle all git operations. This script:

• Parses the repository URL (supports HTTPS, SSH, and short form owner/repo)
• Sets up XDG-compliant directory structure
• Clones the repository (if not present) or updates it (if already cloned)
• Optionally checks out a specific branch
• Returns the repository path

Usage:

# Basic usage (clone/update without branch)
skills/repo-analyzer/scripts/clone_repo.sh --url "$REPO_URL"

# With specific branch
scripts/clone_repo.sh --url "$REPO_URL" --branch "$BRANCH"

Examples:

# HTTPS URL
scripts/clone_repo.sh --url "https://github.com/torvalds/linux"

# Short form (assumes GitHub)
scripts/clone_repo.sh --url "BurntSushi/ripgrep"

# With branch
scripts/clone_repo.sh --url "tiangolo/fastapi" --branch "dev"

What the script does:

• Parses URL to extract platform (github.com, gitlab.com, etc.), org, and repo name
• Creates directory at $XDG_DATA_HOME/claude/repositories/{platform}/{org}/{repo}
• Uses git clone --depth 1 for shallow clones (faster, less disk space)
• Updates existing repos with git fetch and git pull
• Handles authentication via user's git credentials (SSH keys, credential helpers)
• Gracefully continues if pull fails (e.g., detached HEAD, local changes)
• Returns the full repository path to stdout

Step 2: Launch Subagent for Analysis

Now that the repository is ready, launch a general-purpose subagent to analyze it. Pass the repository path and the user's query.

Use the Task tool with subagent_type="general-purpose":

Prompt structure:

Analyze the repository at: {REPO_PATH}

User's query: {USER_QUERY}

Please explore the codebase and provide a comprehensive answer to the user's question.
You have access to all files in the repository.
Use Read, Glob, and Grep tools to navigate and understand the codebase.

The subagent will have full access to read files, search for patterns, and explore the repository structure. It should return findings that directly answer the user's query.

Step 3: Return Findings

Once the subagent completes, present its findings to the user. Include:

• A summary of what was found
• Relevant file paths and line numbers (using the file:line format)
• Any code snippets or examples that answer the query
• The repository location for reference: Repository cloned to: {REPO_PATH}

This gives the user both the answer and the ability to explore the repository themselves if needed.

Example Invocations

Example 1: Analyzing a GitHub repo

User: "Analyze https://github.com/anthropics/anthropic-sdk-python and tell me how to use the messages API"

1. Parse URL → github.com/anthropics/anthropic-sdk-python
2. Clone to ~/.local/share/claude/repositories/github.com/anthropics/anthropic-sdk-python
3. Launch subagent with query: "How to use the messages API"
4. Return findings with examples and file references

Example 2: Checking a specific branch

User: "Look at the 'dev' branch of myorg/myrepo on GitHub and find the new authentication code"

1. Parse URL → github.com/myorg/myrepo
2. Clone/update repo
3. Checkout 'dev' branch
4. Launch subagent with query: "Find the new authentication code"
5. Return findings

Example 3: Short form URL

User: "Investigate django/django and explain how their ORM query optimization works"

1. Assume GitHub → github.com/django/django
2. Clone to ~/.local/share/claude/repositories/github.com/django/django
3. Launch subagent with query: "Explain how ORM query optimization works"
4. Return findings

Error Handling

Handle common failure cases gracefully:

• Invalid URL format: Inform user and ask for a valid Git URL
• Clone failure: Report the git error (e.g., repo doesn't exist, authentication failed)
• Branch doesn't exist: Report error and list available branches if possible
• Subagent timeout: If analysis takes too long, inform user and suggest narrowing the query

Tips for Better Results

• Be specific with queries: "How does authentication work?" is better than "Explain the code"
• Mention specific files or directories if known: "Analyze the src/api/ directory"
• For large repos, narrow the scope: "Focus on the payment processing module"
• If the first analysis isn't enough, the user can ask follow-up questions (repo is already cloned)