| name | sonar |
| description | Claude Code sonar codebase cartographer: workflow guidelines, best practices, instructions, and integration examples |
Sonar Codebase Cartographer
When to activate
Activate when navigating a massive, unfamiliar, or legacy codebase (especially Enterprise monoliths) where reading entire files would quickly exhaust the token context limit. Invoked via /sonar.
When NOT to use
Do not use in very small projects (under 20 files), or when you already know exactly where the code you need to edit is located.
Instructions
- When invoked, your goal is to create a highly condensed "satellite map" of the entire repository containing only the architectural skeletons (Classes, Functions, Structs, and Exports) without the implementation bodies.
- Use the
Bash tool to run extreme filtering commands. For example:
- For TypeScript/JavaScript:
grep -rE '^(export )?(class|const|function|let) .*=' src/ | grep -v node_modules
- For Python:
grep -rE '^(class|def) ' . | grep -v env
- For Go:
grep -rE '^(func|type) ' . | grep -v vendor
(Adapt the regex or use tools like ripgrep or ctags if available on the user's system to get better AST-level extractions).
- Synthesize the raw output into a clean, hierarchical markdown file.
- Save this file to the project root as
CODEBASE_MAP.md.
- Instruct the user: "I have mapped the codebase. From now on, I will consult
CODEBASE_MAP.md first to locate functions before attempting to read massive files, saving you thousands of tokens."
Example
User: /sonar
Claude: Running Sonar mapping...
[Executes grep commands, parsing thousands of files into just their signatures]
[Writes CODEBASE_MAP.md]
I have mapped 4,500 files into a 300-line map. I can now navigate your enterprise monolith without blowing up our context window!
