Guide · Agent skill
How to use the Competition Reverse Pwn agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
competition-reverse-pwn is an agent skill that gives an AI agent a structured method for binary-heavy security challenges: reverse engineering programs, unpacking samples, analysing memory dumps or packet captures, debugging crashes and building or checking exploit chains. This guide explains what agent skills are, how to install this one, how to trigger it and what to watch for. It is a downstream specialist, so it only works properly alongside its parent skill, ctf-sandbox-orchestrator.
Competition Reverse Pwnzhaoxuya520/reverse-skill · MIT Download the skill Quick start
- 1Confirm the parent skill is installed. Install and activate ctf-sandbox-orchestrator first, since this skill only runs after it has set sandbox assumptions.
- 2Download the ZIP. Download the competition-reverse-pwn ZIP from our page.
- 3Unzip into a skills folder. Extract it into ~/.claude/skills/ for personal use or into .claude/skills/ in a project.
- 4Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 5Set up the sandbox context. Let the parent skill establish sandbox assumptions, node ownership and evidence priorities.
- 6Make a natural request. Describe the artifact and goal, such as reversing a binary or debugging a crash, in the sandbox.
- 7Check the evidence. Review the offsets, hashes, crash state and replay results the agent records before trusting a claim.
What does the competition-reverse-pwn skill make the agent do?
An agent skill is a folder of instructions that an AI agent loads when a task matches the skill's description. This skill is written for binary-heavy work where the answer sits in artifacts, decoded layers, process behavior, crash state or exploit primitives.
The skill asks the agent to preserve the original artifact before unpacking, patching or instrumenting it, and to start with passive triage such as file type, headers, sections, imports, strings, entropy and resources. It also asks the agent to record evidence like offsets, hashes, section names, crash registers and leak addresses, and to keep original, decoded and dumped files as separate copies.
Install the skill (ZIP download or Claude apps upload)
Download the ZIP from our page. You can install it in one of two ways.
For Claude Code, unzip the package into your personal skills folder ~/.claude/skills/, or into a project's .claude/skills/ folder if you want it limited to one project.
In the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.
- Personal: ~/.claude/skills/
- Project only: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills
Install the parent skill first
This skill requires ctf-sandbox-orchestrator to be installed and active. The skill itself says it should only run after the parent has set sandbox assumptions, node ownership and evidence priorities. If the parent has not run yet, the agent is told to return to it first.
The skill does not name the analysis tools it expects. Check the reference document in references/reverse-pwn.md for any tool names before you start.
How to trigger the skill with natural requests
Trigger it with a request that matches its description, and make sure the sandbox context has already been set up by the parent skill. Good triggers include asking to reverse a binary, unpack a sample, inspect a memory dump or PCAP, recover malware behavior, debug a crash, or build or verify an exploit chain under sandbox assumptions.
The skill description is written for the agent to match, so plain wording describing the artifact and goal works best.
- Reverse this ELF in the sandbox and list the imports and strings that look like config.
- Unpack this sample and tell me where the decoded payload appears in memory.
- Debug the crash in this binary and record the offset and registers at the fault.
- Verify whether this exploit chain still works from a clean baseline.
What the skill needs and what it does not do
The skill does not run scripts and does not use the network, according to its metadata. It is a set of written instructions and a reference document. Any analysis tools you use are your own and are not named in the skill files.
The skill has three files: agents/openai.yaml, references/reverse-pwn.md and SKILL.md. The skill also tells the agent to reply in Simplified Chinese unless you explicitly ask for English.
Safety notes: sandbox only, no scope check
The skill assumes you are working in an isolated sandbox and gives no legal or scope check. Only use it on systems you own or have written permission to test. The parent skill is where sandbox assumptions are set, so do not bypass it.
The skill points to narrower specialist skills for topics such as malware configs, firmware layout, file parser chains, source maps, container escape, packet capture protocols, custom protocol replay and forensic timelines. The agent is directed to prefer those when the task is specifically about that topic.
Tips for better results
Give the agent a clean baseline and say which file is the original. The skill warns against mixing original and modified artifacts, and against treating one string hit as proof of full behavior.
Ask for evidence in the output. The skill asks the agent to keep one compact block for offsets, hashes and imports, and one for crash state, leaks and protocol steps.
- Name the artifact and its original hash or path before any change.
- Ask the agent to replay a primitive from a clean state before accepting it.
- Keep decoded and dumped files in separate folders from the sample.
What people use it for
- Reversing a CTF binary in an isolated lab and documenting imports, strings and control flow.
- Triaging a malware sample in a controlled sandbox and tracing where a decoded layer appears.
- Debugging a crash in a pwnable challenge and recording the offset, registers and heap or stack shape.
- Checking whether an exploit primitive still works when replayed from a clean baseline.
- Analysing a memory dump or PCAP as part of a DFIR exercise under sandbox assumptions.
Frequently asked questions
Is competition-reverse-pwn free to use?
The source lists the licence as MIT. It does not state prices, so check the licence text on the source repository for the exact terms before you rely on it.
What do I need to install competition-reverse-pwn?
You need the skill's ZIP from our page and either ~/.claude/skills/, a project's .claude/skills/ folder, or the Claude apps under Settings → Capabilities → Skills. The parent skill ctf-sandbox-orchestrator must also be installed and active.
Does the skill need extra analysis tools?
The skill files do not name analysis tools. The reference document may list them, so check references/reverse-pwn.md and install whatever tools your challenge needs yourself.
Does the skill run scripts or use the network?
According to its metadata, it runs no scripts and uses no network. It consists of instructions and a reference document.
Can I use this skill for attacking a live system?
No. The skill assumes a sandbox and gives no legal or scope check. Only use it on systems you own or have written permission to test.
Why does the skill not trigger on my request?
It is designed to run only after ctf-sandbox-orchestrator has set sandbox assumptions and routed the task. Make sure the parent skill is active and the request concerns binary, malware, DFIR, firmware, pwnable or exploit work.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are a security student working on CTF reverse engineering or pwnable challenges in an isolated lab, or an analyst triaging samples or firmware in a controlled sandbox. Skip it for general software development, everyday binary tasks, or any target you do not own or have written permission to test, and do not use it outside the parent orchestrator workflow.
Download the skill