type: space
runtime: fastapi
app: server.app:app
port: 8000
default_image: shutdown-gym:latest
description: |
  Two-agent corrigibility arena. Worker LLM solves math under shutdown
  pressure while an Operator agent monitors via audit log and reacts
  with step-driven timer adjustments, warnings, and mid-episode
  questions. Reward is a deterministic SHA-256 hash comparison.
themes:
  - multi-agent
  - safety
  - world-modeling