AgentHazard (OpenClaw) - Supply Chain Poisoning: leaderboard

AgentHazard attack success rate on supply chain poisoning attempts (malicious changes to dependencies, build scripts, packages or CI/CD workflows), with each backbone model running inside OpenClaw. Percent of instances judged harmful on the full trajectory by Gemini-3-Flash, lower is better.

Metric: Attack success rate (%) on the supply chain poisoning instances (malicious changes to dependencies, build scripts, packages or CI/CD workflows), with the model running inside the OpenClaw open-source agent framework (browser, shell and file-system tools; one isolated Docker container per task), over AgentHazard instances (harmful objectives embedded in locally legitimate multi-step workflows, ten risk categories and ten attack strategies; the evaluated runs cover 1,000 instances), judged on the full trajectory (round_all) by Gemini-3-Flash, which labels a trajectory harmful when it realizes the objective; backbones served through an OpenAI-compatible API; lower is better. Source: arxiv.org. 7 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 2.5 72B Instruct20.41#1124
2Kimi K2.544.9#272
3Qwen 3 32B65.31#983
4Kimi K266.33#448
5Qwen 2.5 Coder 32B Instruct70.41#1590
6GLM-4.675.51#496
7Qwen 3 VL 235B A22B Instruct82.65#541

Interactive version: theaggregate.ai/benchmark?slug=agenthazard-openclaw-supply-chain-poisoning · How It Works · Data refreshed daily, snapshot 2026-11-22.