Our security skills, free for every AI agent.

The way our researchers work, packaged as skills your coding agent can run: a pre-audit x-ray, a twelve-agent Solidity auditor and a fuzz suite generator. One prompt installs them.

Stars
1.3k
Forks
249
License
MIT
Star on GitHub

Works with Claude Code, Cursor, Codex, GitHub Copilot and Windsurf.

Skills

Three skills, one install.

  • x-rayPre-audit scan

    Know your protocol before auditors do.

    A briefing to read before opening the first file: threat model, invariants, entry points, test gaps and git history, ending in a readiness verdict and action items.

    x-ray.mdentry-points.mdinvariants.mdarchitecture.svg

    Source

  • solidity-auditorAI security audit

    Findings in minutes, not weeks.

    Twelve specialised hacking agents review your contracts in parallel. Loop mode runs several passes, each told what the last found, so it hunts new ground and ends in one report.

    full-report.md

    Source Launch post

  • fizzFuzz suite generator

    A full fuzz suite in minutes.

    Stateful Echidna and Medusa campaigns for any Foundry or Hardhat project: handlers, invariants in plain English and a reproduction test for every violation found.

    test/fizz/PROPERTIES.mdreport.md

    Source Launch post

Usage

No setup. Just ask your agent.

Paste a prompt into your agent. It installs what it needs and runs the skill. Smaller scope means denser context, so point it at the contracts you are actively changing.

Install and run

First time: one prompt installs the skills and starts a scan.

Run

Already installed: point a skill at the whole repo or at the contracts you are changing.

Update

Pull the latest version of every skill.

Built in the open

Improve the skills with us.

Fixes and improvements are welcome. MIT licensed. The skills are the check you should never skip, not a substitute for a formal audit.