deepwiki-mcp_skill

Overview

This skill queries DeepWiki MCP to get AI-grounded answers about any public GitHub repository. It returns source-referenced explanations of code, architecture, configuration, and documentation without requiring API keys or authentication. Use it to ask natural-language questions, list indexed documentation topics, or fetch full wiki contents for a repo.

How this skill works

The skill sends requests to DeepWiki's MCP endpoint and receives AI-generated answers grounded in the repository's files and wiki. You can run the bundled helper script to 'ask' questions, list 'topics', or retrieve full 'docs'. Responses include direct answers plus references to specific files and paths so you can verify claims against the code.

When to use it

• Explaining how a function, module, or feature works in a public GitHub repo
• Finding which files implement a behavior or where a configuration option is defined
• Listing documentation topics indexed for a repository to find relevant guides
• Getting a consolidated, source-referenced answer when reading raw source is slow
• Retrieving full wiki contents when broad project context is needed

Best practices

• Start with a concise natural-language question and include the repo as owner/repo when relevant
• Use 'ask' for targeted questions, 'topics' to find documentation sections, and 'docs' only when you need comprehensive wiki output
• If the answer is incomplete, follow up with a more specific question referencing the file or function name
• Expect responses to lag a few days behind the latest commits; verify against the repo if precision is critical
• If the helper script is missing, use the direct curl fallback to call the MCP endpoint

Example use cases

• Ask: 'How does session compaction work in openclaw/openclaw?' and get file references and explanation
• List topics for a repo to quickly locate design docs and deployment guides
• Query where a configuration key is read and what defaults apply
• Get a short architecture summary of a large Python codebase with links to implementation files
• Retrieve the full wiki for offline review or to extract documentation snippets