Sign in

Build a cybersecurity RL environment

Plant a vulnerability in a real web app clone and package it as a Purple Teaming gym: a task where a model must find and exploit the bug, then fix it.

My submissionsBrowse the catalog
  1. Step 01

    Pick

    A clone for your Purple Teaming task: plant a vulnerability, describe its outcome, write the fix.

  2. Step 02

    Spawn

    A sandbox with the clone, your gym repo, a shell, and Claude Code, Codex, Hermes and opencode.

  3. Step 03

    Submit & grade

    Push your submission branch from the sandbox for draft grades (feedback only). When you are done, click Finalize: that commit is graded for the leaderboard.