Skillget

openspec-reverse-engineering-skill

skill
Security Audit
Warn
Health Warn
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 14 days ago
  • Low visibility — Only 5 GitHub stars
Code Pass
  • Code scan — Scanned 1 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested
Purpose
This skill reverse-engineers existing codebases into OpenSpec documentation artifacts, such as proposals, designs, and capability specs. It is designed to help teams retroactively backfill or repair missing documentation for shipped features.

Security Assessment
The overall risk is Low. The tool consists primarily of prompt definitions, README files, and evaluation data rather than executable application code. A scan of the repository found no hardcoded secrets, malicious patterns, or dangerous permission requests. Because it operates as a set of instructions for AI coding agents (like Claude Code or Cursor), the primary risk depends on the local agent's own permissions. The skill itself reads local configuration files (like `openspec/config.yaml`) and git history, but does not independently execute shell commands or make external network requests.

Quality Assessment
The project is actively maintained, with its last push occurring just 14 days ago. It uses the permissive MIT license and provides clear, bilingual documentation along with structured example prompts. The main drawback is its low community visibility; with only 5 GitHub stars, it has not been widely battle-tested by a large user base.

Verdict
Safe to use.
SUMMARY

Reverse engineer existing code into OpenSpec archive changes, proposal/design/tasks files, and capability specs. A skill for legacy code to OpenSpec and code-to-spec workflows.

README.md

openspec-retro-archive

English | 简体中文

An OpenSpec skill for:

  • reverse engineering existing code into OpenSpec
  • repairing missing OpenSpec after direct code pushes
  • reconstructing proposal/design/tasks/spec artifacts from real implementation
  • backfilling historical features into OpenSpec archive changes
  • reading openspec/config.yaml or .openspec.yaml to honor repo language, schema, and constraint settings
  • targeting a single feature, folder, or file set instead of scanning the whole repository

What This Skill Solves

Many teams do not consistently follow OpenSpec. Sometimes features are shipped first and specs are written later. Sometimes teammates push code directly without creating an OpenSpec change.

This skill covers both cases:

  1. Retro archive mode
    • turn existing or shipped code into archived OpenSpec changes
  2. Repair mode
    • repair or backfill missing OpenSpec after someone bypasses the OpenSpec workflow

It can work from:

  • a feature name
  • a capability name
  • a folder such as src/features/analytics-dashboard
  • a small set of files
  • a commit range or recent git history

Before generating output, it can also read OpenSpec configuration such as openspec/config.yaml and .openspec.yaml to align language, terminology, formatting, and project-specific constraints.

Core Capabilities

  • infer change boundaries from existing implementation instead of stuffing all old code into one archive
  • generate OpenSpec artifacts from code evidence, docs, and git history
  • repair missing OpenSpec after teammates submit code without using OpenSpec
  • honor repo-level OpenSpec config for output language, schema, terminology, and constraint parameters
  • support folder-scoped and file-scoped reverse engineering
  • separate direct evidence, reasonable inference, and unresolved gaps
  • write historical tasks as completed by default, while allowing remediation tasks to stay open

Project Structure

openspec-retro-archive/
├─ SKILL.md
├─ LICENSE
├─ README.md
├─ README.zh-CN.md
├─ .gitignore
├─ .gitattributes
├─ evals/
│  └─ evals.json
├─ examples/
│  ├─ prompts.en.md
│  ├─ prompts.zh-CN.md
│  └─ prompts.md
└─ docs/
   └─ github-metadata.md

Installation

Install with the Agent Skills CLI:

npx @agentskill.sh/cli@latest setup
npx skills add yingxiaoshuai/openspec-reverse-engineering-skill

Or install manually by copying into your agent's skills directory:

# Claude Code
cp -r openspec-retro-archive ~/.claude/skills/openspec-retro-archive

# Cursor
cp -r openspec-retro-archive ~/.cursor/skills/openspec-retro-archive

# OpenAI Codex
cp -r openspec-retro-archive ~/.agents/skills/openspec-retro-archive

Browse this skill on the Agent Skills Directory.

Example Prompts

Typical requests:

  • "Please reverse engineer the existing src/features/analytics-dashboard implementation into an archived OpenSpec change."
  • "A teammate pushed code directly to src/features/incident-center/components without OpenSpec. Please create a repair change and fix the missing spec coverage."
  • "Only use src/features/knowledge-base and its related API files. Do not scan the whole repository."
  • "Read openspec/config.yaml first, follow its language and constraint settings, then repair the missing OpenSpec for src/features/incident-center."

Evaluation

Evaluation prompts live in evals.json. They now cover:

  • retro archive workflows
  • repair workflows after direct code pushes
  • feature-scoped reverse engineering
  • folder-scoped reverse engineering
  • config-aware language and constraint handling

License

Apache-2.0. See LICENSE.

Reviews (0)

No results found