emiluzelac

Universal Reverse Engineering — AI skill for Claude Code

AI community

Agent Skill that teaches AI coding agents to reverse engineer how things work: programs, file formats, protocols, devices and processes, evidence-first.

How to install Universal Reverse Engineering

This entry records only its repository, not the path inside it, so there is no exact command to give. Open emiluzelac/universal-reverse-engineering and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Universal Reverse Engineering does

Agent Skill that teaches AI coding agents to reverse engineer how things work: programs, file formats, protocols, devices and processes, evidence-first. Works in Claude Code, Qwen Code, Codex and any Agent Skills host.

Alternatives in AI

Agency Agents — A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy inject 148k ★
Pretty Mermaid Skills — To provide AI with Mermaid chart rendering capability, supporting both SVG and ASCII output formats 552 ★
Update Branch Name — by giselles-ai - Updates branch names with proper prefixes and formats, enforcing naming conventions, supporti 502 ★

README

![Universal Reverse Engineering Skill](assets/ure-header.jpg)

Universal Reverse Engineering Skill

[![validate](https://github.com/emiluzelac/universal-reverse-engineering/actions/workflows/validate.yml/badge.svg)](https://github.com/emiluzelac/universal-reverse-engineering/actions/workflows/validate.yml)

An [Agent Skill](https://agentskills.io/specification) that teaches an AI coding agent to work out **how things work**: a program, a file format, a protocol, a gadget, a product, a process. It is evidence-first. The agent observes, builds a model, tests it, and explains, and it keeps *observed*, *inferred*, *verified* and *unknown* apart the whole way.

It is one folder, `reverse-engineering/`, in the open Agent Skills format, so the same copy works in Claude Code, Qwen Code, Codex and any other host that reads that format. It also degrades gracefully: pasted into a plain chat window with no tools, it still runs the method without inventing tool calls.

The skill's own [README](reverse-engineering/README.md) has the full install table, a first hands-on test, notes for bare chat endpoints, and the version history.

Install

git clone https://github.com/emiluzelac/universal-reverse-engineering.git
cd universal-reverse-engineering
mkdir -p ~/.claude/skills ~/.agents/skills
cp -R reverse-engineering ~/.claude/skills/     # Claude Code
cp -R reverse-engineering ~/.agents/skills/     # Qwen Code and Codex

Restart the agent, then type `/reverse-engineering` (Claude Code, Qwen Code) or `$reverse-engineering` (Codex), or just ask how something works. Each host also loads the skill on its own when a request matches its description.

What is in the skill

Two modes. A quick explanation for conceptual questions and photos, and a full investigation with an investigation plan, a hypothesis ledger and a run log when there are artifacts to test. The agent picks the mode first.
The loop. Observe, decompose, hypothesize, test, reconstruct, verify,