repomix-explorer

Repomix Explorer is a codebase-analysis skill that helps users understand an unfamiliar or large repository, remote or local, by packing it with the Repomix CLI and then reading and searching the generated output. It is aimed at high-level comprehension rather than targeted edits: getting a structural overview, discovering patterns across many files, or gathering metrics like file and token counts and language ratios.

The workflow packs the repo with npx repomix@latest, using --remote for GitHub repos (always writing the output to /tmp to avoid polluting the user's project) or a local directory path, and choosing an output format (XML by default, with markdown, JSON, or plain available) and optional Tree-sitter --compress for large repos. It then checks the reported metrics, searches the packed output with grep patterns for functions, classes, imports, config, auth, API endpoints, models, and error handling, and reads specific sections with offset/limit before summarizing findings and suggesting next steps. The skill is explicit about what NOT to trigger on: editing or refactoring the current project, reading a known local file (use Read or grep directly), single-symbol lookups, or git/test/build/install operations. It offers optional cleanup of large output files after analysis.

Target users are developers, reviewers, and technical writers who need a fast, read-only orientation to a codebase, especially remote GitHub repositories they want to explore before diving in. Its operations are read-only analysis: it packs, greps, and reads, and only deletes the temporary Repomix output file it created.