name: YC-Bench
description: >
  Long-horizon agent benchmark where an LLM plays CEO of an AI startup
  over a simulated 1-year run via CLI tool use. Tests sustained strategic
  coherence: resource management, deadline reasoning, prestige climbing,
  and adversarial client detection.
evaluation_framework: yc-bench

tasks:
  - id: medium