What is an As-Is specification?

An As-Is specification describes how the code currently functions based on its actual implementation, rather than how it was originally intended to work (To-Be).

Why is human review still necessary for Layer C?

Layer C covers business meaning and intent; while Claude can infer purpose from naming and logic, true business context often resides in human knowledge outside the source code.

Does it handle code changes and regressions?

Yes, it includes a diff detection workflow that compares git changes against existing specs to identify which parts of the documentation need updating.

Can it generate API documentation?

Yes, it extracts interface details such as request/response schemas, paths, and status codes in OpenAPI-compatible YAML format.

How does Spec-Gen ensure documentation accuracy?

It uses a confidence labeling system (Verified, Observed, Assumed) and provides specific file/line evidence for every generated specification so developers can verify the source.

Spec-Gen Code Documentation

Name: Spec-Gen Code Documentation
Author: CAPHTECH

byCAPHTECH

0•

学习与文档

Extracts structured, evidence-based 'As-Is' specifications from source code across interface, behavior, and business logic layers.

Spec-Gen is a specialized Claude Code skill designed to automate the reverse-engineering of code into formal documentation. By analyzing source code and tests, it generates accurate 'As-Is' specifications categorized into three layers: technical interfaces, logical behaviors, and business intent. Unlike standard documentation generators, Spec-Gen includes confidence labels (Verified, Observed, Assumed) and direct code evidence links for every claim, making it an essential tool for legacy code audits, API documentation, and identifying discrepancies between current implementation and intended behavior.

主要功能

01Three-layer specification extraction covering technical interfaces, logical behaviors, and business intent.

02Evidence-based verification with direct file paths and line number references for every claim.

03Granular confidence labeling (Verified, Observed, Assumed, Unknown) to distinguish proven facts from logical inferences.

040 GitHub stars

05Support for multiple entry points including HTTP APIs, CLI tools, event subscribers, and public library exports.

06Automated diff detection to identify how recent code changes impact existing specifications and documentation.

使用场景

01Detecting discrepancies between documentation and actual code behavior during refactoring or audits.

02Generating OpenAPI-compatible API schemas directly from existing backend implementation code.

03Reverse-engineering legacy systems to create accurate, up-to-date technical documentation.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add caphtech/claude-marketplace spec-gen

For use in Claude.ai and ChatGPT

主要功能

01Three-layer specification extraction covering technical interfaces, logical behaviors, and business intent.

02Evidence-based verification with direct file paths and line number references for every claim.

03Granular confidence labeling (Verified, Observed, Assumed, Unknown) to distinguish proven facts from logical inferences.

040 GitHub stars

05Support for multiple entry points including HTTP APIs, CLI tools, event subscribers, and public library exports.

06Automated diff detection to identify how recent code changes impact existing specifications and documentation.

使用场景

01Detecting discrepancies between documentation and actual code behavior during refactoring or audits.

02Generating OpenAPI-compatible API schemas directly from existing backend implementation code.

03Reverse-engineering legacy systems to create accurate, up-to-date technical documentation.