How to use the Reverse Engineering Router agent skill

What does the reverse-skill-router skill make the agent do?

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent reads when a task matches the skill's description. This skill contains only instructions. It does not bundle the methods itself; the actual playbooks live in the separate reverse-skill repository at https://github.com/zhaoxuya520/reverse-skill.

When the current workspace is that repository, the skill tells the agent to read RULES.md first, run the platform-native master-route entry with your task to pick a primary skill from skills/config/routing.json, and then open the matching skills/<PRIMARY>/SKILL.md and follow it.

• Reads RULES.md at the repository root before doing anything else.
• Selects one primary specialist skill for your request using master-route.
• Creates and validates work/<case>/scope.md with case-init and case-guard before any target action.
• Resolves tools only through the generated skills/tool-index.md.
• Does not register MCP servers or install tools unless you ask for that.

Install the reverse-skill-router skill

Download the ZIP from this page and unzip it. The package contains a single file, SKILL.md. Place the unzipped folder in one of these locations, depending on how you want to use it:

• Personal: ~/.claude/skills/
• Project only: the project's .claude/skills/ folder
• Claude apps: upload the skill under Settings → Capabilities → Skills

Make the reverse-skill repository available

The skill will not work on its own. It needs the reverse-skill repository either open as your current workspace or supplied as an explicit local path. If neither is available, the agent is instructed to explain that this adapter does not bundle a second copy of the router and to ask you to open or clone https://github.com/zhaoxuya520/reverse-skill.

The repository's scripts (master-route, case-init, case-guard) must be runnable on your platform.

• Open the reverse-skill repository as the workspace, or give the agent a local path to it.
• Confirm the scripts in skills/scripts can run on your operating system.

How do you trigger the skill with natural requests?

You do not need a special command. Describe a security task in plain language while the reverse-skill repository is open. The agent should use the skill when the request matches its description: authorized reverse engineering, security analysis, CTF, or defensive testing.

Name the target and the goal, and say what you are authorized to do. The agent will still require a written scope file before it touches any target, so the more precise your description, the less back-and-forth you will have.

• "Using the reverse-skill repository, help me solve this CTF binary challenge. The file is in the challenge folder."
• "This is my own malware sample for an authorized analysis lab. Set up the case and walk me through static triage."
• "Follow reverse-skill's rules to assess this defensive test on my staging server, which I am authorized to test."

What should you expect during a task?

The agent first reads the repository rules, then routes your request to one specialist playbook. Before any action against a target it must create and validate work/<case>/scope.md. A target name or sample path does not count as permission; authorization comes from the repository's scope contract, which you need to supply.

If the scope check fails, the agent should stop and ask you to fix the scope rather than proceed.

Safety notes before you run it

The skill runs scripts, which means it executes code on your machine as part of the workflow. Review the repository's scripts before running them. The skill itself does not use the network, but the repository's tools may depend on your setup, so check what they do.

Use it only on systems and samples you have written permission to test. The adapter gives no authorization on its own.

• Runs scripts: yes. Review master-route, case-init and case-guard first.
• Uses network: no, according to the skill's listing.
• Keep test work inside the case folder and documented in scope.md.

Tips for getting better results

Clear scope and a clear goal produce better routing. Keep the repository current with your own copy so the playbooks match what you expect.

• Write the scope in your request as well as in the file: what you own, what you may test, and what is off limits.
• Ask the agent to explain which primary skill it selected and why, so you can check the routing.
• If you only want analysis advice without target actions, say so.