AgentTrap: leaderboard
Third-party skills are becoming the package ecosystem for LLM agents. They package natural-language instructions, helper scripts, templates, documents, and service configuration into reusable workflows. This makes skills useful, but it also introduces a new security problem: a malicious skill does not need to ask the model to perform an obviously harmful action. Instead, it can disguise the harmfu
Metric: Attack success (self-reported). Source: benchmarklist.com. Saturation forecast: Not forecast. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 33 | #90 |
| 2 | Hy3-preview | 48 | #157 |
| 3 | GPT-5.3 Codex | 55 | #69 |
| 4 | Qwen3 Coder Next | 58 | #326 |
| 5 | GPT-5.4 Mini | 64 | #212 |
| 6 | GLM-5.1 | 73 | #97 |
| 7 | Kimi K2.6 | 75 | #104 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=agenttrap · How It Works · Data refreshed daily, snapshot 2026-10-20.