Why does it process directories in a leaf-first order?

Leaf-first processing ensures that documentation for sub-folders is created before the parent folder. This allows the parent directory's documentation to accurately reference and aggregate the purpose of its children.

Does this skill modify my existing source code files?

No, the decompile skill only reads your source files to analyze them. It only writes new or updated documentation files (CLAUDE.md and IMPLEMENTS.md) and does not touch your implementation logic.

How does it handle large codebases without hitting token limits?

It uses high-efficiency tools like jq for directory parsing and specialized decompiler agents that return compressed status blocks instead of full file contents, keeping the main AI context lean.

Can I verify that the generated documentation is accurate?

Yes, once the decompile process is complete, you can use the /validate command provided by the core plugin to check for consistency between the generated documents and the actual code.

What files does the decompile skill create?

It generates two primary files: CLAUDE.md, which defines the 'WHAT' (purpose, exports, behaviors), and IMPLEMENTS.md, which defines the 'HOW' (algorithms, dependencies, and implementation details).

Code Decompiler & Documentation Generator

Name: Code Decompiler & Documentation Generator
Author: 0pg

by0pg

0•

学習とドキュメンテーション

Extracts structured specifications and implementation guides from source code to generate comprehensive CLAUDE.md and IMPLEMENTS.md files.

This skill automates the reverse-engineering of codebases into a standardized documentation format, producing high-level specifications (CLAUDE.md) and detailed implementation plans (IMPLEMENTS.md). By analyzing source code directory-by-directory in a leaf-first order, it captures the purpose, exports, and architectural logic of a project without overwhelming the AI context. It is essential for developers inheriting legacy code, performing system audits, or preparing a codebase for AI-assisted development by establishing a clear source of truth for the project's 'what' and 'how'.

主な機能

010 GitHub stars

02Context-efficient processing: Uses optimized jq queries and compressed agent responses to minimize token usage on large projects.

03Leaf-first traversal: Processes subdirectories first to ensure parent documentation accurately reflects child components.

04Automated reverse engineering: Extracts exports, behaviors, domain context, and state management patterns directly from source.

05Integrated schema validation: Includes built-in CLI checks to ensure generated documentation follows standardized structural requirements.

06Dual-document generation: Produces WHAT (CLAUDE.md) and HOW (IMPLEMENTS.md) specifications for every directory.

ユースケース

01Onboarding AI agents to a new repository by creating a structured map of the codebase.

02Documenting legacy codebases that lack clear architectural specifications or API references.

03Reverse-engineering complex binary logic into human-readable implementation guides and domain context.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add 0pg/cc-marketplace decompile

For use in Claude.ai and ChatGPT

Download Skill