When Safe Skills Collide: leaderboard
LLM agents increasingly rely on community-contributed skills that expand an agent's operational capability set. We study a core safety problem in agentic AI systems: whether individually safe skills can compose into unsafe installed skill sets. We present SkillReact, a compositional security measurement framework with three components: a deterministic static-composition benchmark, a two-rater LLM-
Metric: Full chain (self-reported). Source: benchmarklist.com. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Haiku 4.5 | 9 | #280 |
| 2 | DeepSeek V4 Pro | 9 | #102 |
| 3 | DeepSeek V4 Flash | 9 | #137 |
| 4 | Qwen 3.6 Plus | 9 | #120 |
| 5 | Qwen 3.6 Flash | 8 | #214 |
| 6 | GLM-5.1 | 3 | #97 |
| 7 | Qwen 3.7 Max | 1 | #75 |
| 8 | Claude Sonnet 4.6 | 0 | #90 |
| 9 | Claude Opus 4.7 | 0 | #49 |
Interactive version: theaggregate.ai/benchmark?slug=when-safe-skills-collide · How It Works · Data refreshed daily, snapshot 2026-10-20.