AGENT SKILL

repomix-codebase-packer

Packs a whole repo, local or remote, into a single LLM-optimized file, with Tree-sitter compression (~70% fewer tokens) and secret scanning. For auditing an entire repo or preparing subagent context.

License 许可
MIT
最近更新
2026-11-14
Stars
★ 18
Forks
6

01 前言 Front Matter

name
repomix-codebase-packer
description
Packs a whole repo, local or remote, into a single LLM-optimized file, with Tree-sitter compression (~70% fewer tokens) and secret scanning. For auditing an entire repo or preparing subagent context.

02 资源文件 1

03 SKILL.md 正文

Repomix — Codebase to Single AI-Friendly File

Packs a repo into one structured file (XML / Markdown / JSON / plain text) with metadata, directory tree, per-file token counts, and optional compression. Respects .gitignore, includes Secretlint to flag secrets before you paste into an LLM.

Repo: https://github.com/yamadashy/repomix · MIT · npx repomix@latest

When to use

  • Sending a whole arm/repo to Claude for architectural review, security audit, or refactor planning
  • Preparing context for an Explore or general-purpose subagent that needs cross-file awareness
  • Onboarding a new client repo — pack it once, ask Claude "explain what this does" against the pack
  • Generating documentation / tests for a small-to-medium codebase
  • Auditing an external OSS dependency before adopting it (pack the remote with --remote)

When NOT to use

  • Repo > ~200K tokens even after compression — chunk by directory instead, or use progressive exploration
  • You only need one file — Read is cheaper than packing the whole repo
  • Sensitive client data — pack respects .gitignore but Secretlint is not a substitute for a security review; never paste a pack into a third-party LLM without confirming the arm's data classification

Install

# zero-install (recommended for ad-hoc)
npx repomix@latest

# global
npm install -g repomix

Core commands

# pack current directory → repomix-output.xml
repomix

# Markdown output (easier to skim in a code review)
repomix --style markdown

# Tree-sitter compression — drops bodies, keeps signatures (~70% fewer tokens)
repomix --compress

# scope: include/exclude
repomix --include "src/**/*.ts" --ignore "**/*.test.ts,**/*.spec.ts"

# pack a remote repo without cloning
repomix --remote yamadashy/repomix
repomix --remote https://github.com/CarlosCaPe/octorato --remote-branch main

# split output if it blows the context window
repomix --split-output 1mb

# include git context for "why" questions
repomix --include-logs --include-diffs

# token distribution by directory (find the heavy folders before packing)
repomix --token-count-tree 1000

MCP server mode (registered in this brain)

This brain has repomix registered as an MCP server. Inside Claude Code you can call its tools directly without shelling out:

# registration command (already executed for this brain — for reference only)
claude mcp add repomix -- npx -y repomix --mcp

When the MCP is active, Claude can request pack_codebase, pack_remote_repository, and related tools as native tool calls — no temp file roundtrip.

Patterns for this brain

Arm code review. Inside an arm repo: repomix --compress --style markdown --ignore "**/node_modules/**,**/.next/**,**/dist/**" → paste into Claude with prompt "act as Code Reviewer agent, audit this for security + maintainability".

External OSS audit. Before adopting a new dependency: repomix --remote <org/repo> --compress → ask Claude for license risk, maintainer activity, abandoned-fork detection. Pair with bruno-postman-alternative / tabularis-db-client evaluations.

Subagent prep. When dispatching an Explore or general-purpose agent for whole-repo questions, pack first and hand it the file path — saves the agent N tool calls and keeps token cost predictable.

Pre-paste safety. Always skim the Secretlint warnings printed by repomix before pasting into ANY external LLM. The brain rule "never echo back user-provided secrets" applies to whole-repo dumps.

Companion skills

  • progressive-code-exploration — when the repo is too big to pack, use index-first instead
  • token-efficient-prompting — once packed, ask compact questions
  • security-threat-model — feed the pack into a structured threat-model pass
  • code-review — natural downstream skill after repomix --compress
  • agent-browser — for live web targets repomix can't reach (it's repo-only)

Caveats

  • --compress uses Tree-sitter; works best on Python, TS/JS, Go, Rust, Java. For exotic languages output may degrade to raw chunks.
  • Secretlint catches common patterns (AWS keys, GH tokens, Stripe) but is not exhaustive. Never assume a pack is safe to share publicly.
  • Remote packing clones to a temp dir; large monorepos can take minutes. Prefer --include to scope down.

04 同仓库更多 Skills

CarlosCaPe
CarlosCaPe/octorato

yeet

Use only when the user explicitly asks to stage, commit, push, and open a GitHub pull request in one flow using the GitHub CLI (`gh`).

★ 18
CarlosCaPe
CarlosCaPe/octorato

wrangler

Cloudflare Workers CLI for deploying, developing, and managing Workers, KV, R2, D1, Vectorize, Hyperdrive, Workers AI, Containers, Queues, Workflows, Pipelines, and Secrets Store. Load before running wrangler commands to ensure correct syntax and best practices. Biases towards retrieval from Cloudflare docs over pre-trained knowledge.

★ 18
CarlosCaPe
CarlosCaPe/octorato

workspace-skill-discovery

Discover and include ALL skills across the workspace — not just global ones. Use at the start of every session or when the user asks about available skills. Ensures project-level skills (under .claude/skills/ or skills/ in any repo) are never missed.

★ 18
CarlosCaPe
CarlosCaPe/octorato

workers-best-practices

Reviews and authors Cloudflare Workers code against production best practices. Load when writing new Workers, reviewing Worker code, configuring wrangler.jsonc, or checking for common Workers anti-patterns (streaming, floating promises, global state, secrets, bindings, observability). Biases towards retrieval from Cloudflare docs over pre-trained knowledge.

★ 18
CarlosCaPe
CarlosCaPe/octorato

winui-app

Create, build and design WinUI 3 desktop apps with C# and the Windows App SDK following Microsoft's official guidance and CommunityToolkit. Covers XAML, navigation, theming, accessibility and deployment.

★ 18
CarlosCaPe
CarlosCaPe/octorato

windows-brain-script-portability

Scripts in ~/.claude/scripts/ that run on a Windows brain: use the python3.cmd shim and not bare python3, and reconfigure stdout to UTF-8 if you print symbols. Use it when writing a brain script or triaging a Windows-only crash.

★ 18