Develop a working exploit for a security challenge with agents
A security professional set up AI agents to develop an exploit for the brotli2bitcoin challenge, which took 37 minutes and won the challenge. ·
Done with a custom-built or unnamed agent
- What they did
- A security professional configured AI agents to replicate the same manual exploitation approach they would normally use themselves against the brotli2bitcoin.com challenge, letting the agents carry out the steps independently.
- How it went
- The exploit succeeded and won the challenge in 37 minutes, which the author estimates would cost around $11 in API tokens, though with their existing subscription the real cost was just a few cents.
- Worth knowing
- Setting up the agents required real expertise first, mirroring the author's own manual methodology rather than fully automating from scratch.
Try it yourself with Claude Code
I'm working on a security CTF challenge called [challenge name/description]. Help me analyze the target, identify the vulnerability, and iteratively develop and test a working exploit script, running it against the challenge until it succeeds. Explain each step of the exploit so I understand how it works, and stop before submitting or sending anything to an external server until I confirm it's ready.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: HN · Nov 21, 2026
Related use cases
Find a Linux vulnerability hidden for 23 years
An account of Claude Code uncovering a long-standing vulnerability in Linux.
Generate parametric geometry workflows with an agent
GrassLobster uses agentic AI to generate parametric geometry workflows.
Custom agent
Reverse engineer a patch into a full exploit chain
A researcher used Claude Opus in Claude Code with Playwright MCP to decompile code, find vulnerabilities, and build a working exploit.
Hunt for unpatched vulnerabilities in SQLite commits
Google's Big Sleep agent reviewed recent SQLite commits starting from a patched bug and found an exploitable flaw, which was fixed before release.
Reverse engineer a Go binary with Ghidra via MCP
Huli had Cursor drive GhidraMCP to identify a Golang binary, list its routes and find a path traversal, but it missed the SQL injection check.
Mine biological datasets for new leads with thousands of agents
OpenAI describes running thousands of AI agents over complex biological data to uncover new research leads.
Custom agent
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.