repomix_skill

Overview

This skill packages entire repositories into single, AI-friendly files (XML, Markdown, JSON, or plain text) to prepare codebases for LLMs or human review. It streamlines token-aware snapshots, comment-stripping, and security scanning so you can feed focused context to models like Claude, ChatGPT, or Gemini. Use it to create reproducible, shareable repository contexts for audits, documentation, and debugging.

How this skill works

Repomix walks the repository tree (local or remote), applies include/ignore patterns and .gitignore rules, and converts files into a single formatted output with clear separators and token counts. It can strip comments, run Secretlint-style security checks, and produce token-count trees so you can target files that fit your model’s context window. Outputs are generated in multiple styles (markdown, xml, json, plain) and can be copied to the clipboard or written to a file.

When to use it

• Prepare a codebase snapshot for LLM analysis or pair programming
• Package a remote third-party library for a security audit without cloning locally
• Create context for documentation generation or automated summarization
• Investigate cross-file bugs by bundling relevant modules into one file
• Optimize inputs to fit a specific LLM token limit

Best practices

• Run security checks and review output before sharing externally
• Use --include and --ignore patterns to limit token-heavy files
• Generate a token-count tree to identify large contributors and trim or compress them
• Strip comments for cleaner LLM inputs when comments are not needed
• Add a .repomixignore for sensitive files and avoid packaging .env or private keys

Example use cases

• Package a feature branch for an AI code review: include src/**, remove comments, output markdown
• Audit a vendor library remotely: npx repomix --remote owner/repo --style xml -o audit.xml
• Produce documentation context by bundling code and docs: repomix --include "src/**,docs/**,*.md" --style markdown
• Investigate a bug across services: repomix --include "src/auth/**,src/api/**" -o debug-context.xml
• Manage LLM context: repomix --token-count-tree and then use include patterns to fit your target model