How does the --tech-agnostic flag work?

When used, the skill describes the project's needs in abstract terms (e.g., 'persistent relational storage') rather than naming specific technologies like PostgreSQL or Prisma, making the spec portable across different stacks.

What is a SPEC.md file?

A SPEC.md is a portable requirements specification that describes what a project does (features, flows, data entities) without implementation details, serving as a blueprint for rebuilding or documenting the software.

Can I use the output to rebuild my project with AI?

Yes. The SPEC.md is specifically designed to be high-level enough that a competent AI agent can use it as a 'source of truth' to rebuild the entire application with cleaner, more modern code.

Does this skill read every file in my repository?

No. It uses a smart analysis workflow that first reads the directory tree and then selectively analyzes representative files like manifests, routes, and schemas to be as token-efficient as possible.

Project Spec Extractor

Name: Project Spec Extractor
Author: tsilva

bytsilva

0•

학습 및 문서화

Analyzes your codebase to generate a clean requirements specification that captures what a project does without implementation details.

Project Spec Extractor is a specialized Claude Code skill designed to reverse-engineer project requirements from source code. It follows a rigorous four-phase analysis—Identity, Feature Discovery, Data & Integrations, and Behavioral Verification—to produce a standardized SPEC.md file. This output serves as a portable blueprint, allowing developers or AI agents to rebuild projects with cleaner code while maintaining identical functionality. Whether you're documenting a legacy system or preparing for a tech stack migration, this skill distills complex repositories into clear, actionable requirements.

주요 기능

01Structured four-phase analysis (Identity, Features, Data, and Verification)

02Behavior-centric extraction that focuses on 'what' rather than 'how'

03Standardized 9-section SPEC.md output including user flows and acceptance criteria

040 GitHub stars

05Tech-agnostic mode to describe abstract capabilities instead of specific stacks

06Smart, token-efficient file scanning that prioritizes manifests and entry points

사용 사례

01Automating high-quality documentation for undocumented or inherited repositories

02Recovering requirements from legacy codebases for modern rewrites

03Generating portable blueprints for AI agents to rebuild projects from scratch

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tsilva/claudeskillz project-spec-extractor

For use in Claude.ai and ChatGPT

Download Skill