Add metadata and links to ClawProBench dataset card

README.md

@@ -1,4 +1,32 @@
 ---
 license: apache-2.0
+task_categories:
+- other
+tags:
+- agent
+- benchmark
+- runtime
 ---
-Cite arxiv.org/abs/2608.22510 
+
+# ClawProBench
+
+ClawProBench is a trace-aware benchmark for runtime-native agent evaluation, instantiated on OpenClaw. It provides a live-first evaluation harness with 102 active scenarios and a frozen holdout set for robust ranking. Trials are scored from execution traces via a safety-gated formula combining correctness, process quality, and efficiency.
+
+- **Paper**: [ClawProBench: Trace-Aware Evaluation of AI Agents with Runtime Coverage and Frozen Workplace-Style Holdouts](https://arxiv.org/abs/2608.22510)
+- **Project page**: [https://suyoumo.github.io/bench/](https://suyoumo.github.io/bench/)
+- **GitHub repository**: [https://github.com/suyoumo/ClawProBench](https://github.com/suyoumo/ClawProBench)
+
+## Sample usage
+
+A one-trial smoke run on the default ranking benchmark:
+
+```bash
+python3 run.py run \
+  --model '<MODEL>' \
+  --execution-mode live \
+  --benchmark-profile core \
+  --trials 1 \
+  --cleanup-agents
+```
+
+For full details, see the [GitHub repository](https://github.com/suyoumo/ClawProBench).