cc-rsg-web: Reverse-Generate Specs from Code — Turning Legacy Codebases into Maintainable Assets with Claude Code

Legacy code maintenance sites often lack specification documents, creating significant burdens during handovers and modifications. cc-rsg-web, released on Zenn, is a web application that automatically generates draft specifications from codebases. This tool leverages Claude Code to turn legacy systems into maintainable assets.

📑Table of Contents

1. What is cc-rsg-web — A Web App for Reverse-Generating Specs from Code
2. Design Philosophy and the 6+1 Phase Reverse-Generation Flow
3. Supported Languages, Templates, and Question Bank Features
4. Practical Use Cases and Benefits
5. Installation and Getting Started
6. Frequently Asked Questions (FAQ)
7. Summary

---

What is cc-rsg-web — A Web App for Reverse-Generating Specs from Code

cc-rsg-web is a web application that reverse-generates specification documents from existing code. It addresses the challenge of missing specs in legacy codebases and is offered as a browser-accessible version of the cc-rsg OSS. The target users are maintenance engineers and handover personnel. Details are covered in the Zenn article at https://zenn.dev/daishiro/articles/cc-rsg-web-release, positioning it as the web edition of cc-rsg OSS.

In real development sites, past code often remains without documented design intent or constraints. This tool analyzes such code and produces base documents for specifications. Being accessible via browser, it requires no special setup.

---

Design Philosophy and the 6+1 Phase Reverse-Generation Flow

The design philosophy of cc-rsg-web emphasizes honesty, traceability, and prevention of omissions. The reverse-generation process consists of seven phases from Phase 0 to Phase 6, each serving a specific role.

Phase 0 defines goals to clarify the overall purpose. Phase 1 (Recon) grasps the overall picture of the code, Phase 2 creates a WBS to break down the work. After investigation in Phase 3, verification in Phase 4, and interactive refinement in Phase 5, delivery occurs in Phase 6.

There are three Depth modes: comprehensive, outline, and interactive, selectable according to purpose. Comprehensive prioritizes details, outline focuses on overview, and interactive proceeds in dialogue form. Output artifacts include state.json, recon-report.md, inventory, wbs, and final spec, all managed traceably.

This flow ensures that generated specifications explicitly state their basis, making later confirmation and correction easy. The Zenn article also explains this Phase structure in concrete terms.

---

Supported Languages, Templates, and Question Bank Features

cc-rsg-web supports multiple stacks including Ruby on Rails, Python, and JavaScript/TypeScript. The granularity of inventory is defined, systematically extracting code structure.

A Question Bank consisting of seven standard categories is prepared, with questions systematized to resolve uncertainties. Each answer is assigned a Confidence label (VERIFIED, INFERRED, ASSUMED) to distinguish inference from facts.

Templates are standardized to maintain consistent output quality. This allows specification documents in a consistent format across different projects. The source code is published in the GitHub repository at https://github.com/daishir0/cc-rsg, where detailed implementations can be checked.

---

Practical Use Cases and Benefits

Use cases for cc-rsg-web include modernization of legacy systems. By analyzing old codebases and reconstructing specifications, migration plans become easier to formulate.

It is also effective for onboarding new engineers. The overall picture and specifications of the code can be grasped in a short time, reducing learning costs. It can also be used for preparing specifications as deliverables, proving particularly valuable in projects with insufficient documentation.

Compared to manual creation, there is a significant difference in man-hours and quality. Work that takes several weeks manually can generate a draft in a few hours. In terms of quality, the ensured traceability and explicit inference sections are superior.

Tool	Input	Output Format	Traceability	Inference Explicit	Resumability
cc-rsg-web	Codebase	Markdown spec	[REF: path:line]	Dedicated chapter	state.json
Manual Creation	None	Free	Arbitrary	None	None
General LLM	Code snippet	Free form	Weak	Weak	None

Source: Zenn article (as of June 2026) and GitHub repository

---

Installation and Getting Started

Installation is done from the GitHub repository. Clone the repository and install the required dependencies. Skill placement for Claude Code is completed simply by placing files in the specified directory.

The basic command flow involves starting the web app, specifying the codebase path, and beginning analysis. The output directory generates various files including state.json to manage progress.

Detailed procedures are described in the Zenn article and the README of the GitHub repository. Since it can be operated in the browser, it can be tried immediately if the Claude Code environment is ready.

---

Frequently Asked Questions (FAQ)

What is the difference between cc-rsg-web and cc-rsg?
cc-rsg-web is the web version of cc-rsg OSS and can be used from a browser. While cc-rsg focuses on local execution, the web version facilitates remote access and sharing.

What if the language or framework is not supported?
Extend the Question Bank or supplement manually. Basic structure extraction is possible even for unsupported cases.

How are inferred parts handled (undetermined matters chapter)?
They are summarized in a dedicated chapter and distinguished by Confidence labels. Undetermined matters can be verified later.

Can I resume even if I interrupt the session?
You can resume from the interruption point using state.json. Session management is built-in.

What about commercial use or enterprise adoption licenses?
Check the license in the GitHub repository. It is published as OSS, with conditions for commercial use described.

What is the quality of the generated specification documents?
A certain level of quality is guaranteed by the Phase structure and Confidence labels. Manual review is recommended.

What is the difference from other AI tools (ChatGPT, Gemini)?
cc-rsg-web targets the entire codebase and performs reverse generation with a Phase structure. Traceability and resumability are its features.

---

Summary

cc-rsg-web is a tool that efficiently generates specifications from legacy code. This web app, released on Zenn, contributes to solving challenges in development sites. Refer to the GitHub repository and try it in practice. As a next action, we recommend applying cc-rsg-web to your own projects and advancing specification maintenance.