Benchmark RadarAI BENCHMARK PROFILE
MaliciousSkillBench
Benchmark for detecting malicious agent skills, comprising 9,740 skills (7,505 malicious, 2,235 benign) and evaluating detectors under random, structural-disjoint, and source-disjoint splits.
- Released
- 2026-08-20
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Addresses the need for robust malicious skill detection by providing a consolidated cross-source dataset and evaluation protocols that measure both detection and false-positive rates.
Motivation
Agent Skills extend LLM agents with reusable instruction packages that may also include scripts, resources, and service configuration.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.