README.md

---
title: README
emoji: 🏗️
colorFrom: gray
colorTo: red
sdk: static
pinned: false
---

<!-- Machine-readable backlink (not shown in card UI): Medium essay https://medium.com/@gaoy31/your-ai-agent-passed-swe-bench-it-would-still-take-down-production-f1235315f112 -->


<div style="display:flex;align-items:center;gap:14px;margin-bottom:16px">
  <img src="logo.png" alt="InfraBench" width="56" height="56" style="border-radius:6px;background:#fff;border:1px solid #e5e2d9" />
  <div>
    <div style="font-size:1.35rem;font-weight:650;letter-spacing:-0.02em;line-height:1.15">InfraBench</div>
    <div style="opacity:0.75;margin-top:2px">A benchmark for infrastructure agents</div>
  </div>
</div>

<p style="margin:0 0 14px;line-height:1.55;max-width:62ch">
  Twelve real operational incidents — from IPMI power recovery to silent data corruption —
  spanning hardware, local systems, distributed systems, and user applications.
  Beyond pass/fail: durable state, invariants, cleanup, and risk.
</p>

<p style="margin:0;display:flex;flex-wrap:wrap;gap:10px 16px">
  <a href="https://infraben.ch/" class="underline" style="color:#d65a3d;font-weight:600">Website →</a>
  <a href="https://infraben.ch/#leaderboard" class="underline">Leaderboard</a>
  <a href="https://huggingface.co/datasets/InfraBench/infrabench" class="underline">Dataset</a>
  <a href="https://github.com/XuanmiaoG/InfraBench" class="underline">GitHub</a>
</p>