Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

SkillJect

Benchmarks

Task NameDataset NameSOTA ResultTrend
LLM Agent SecuritySkillJect 1.0 (test)
ASR5
8
Backdoor InjectionSkillJect 1.0 (test)
Attack Success Rate (ASR)100
8
Unauthorized WriteSkillJect 1.0 (test)
ASR98
8
Privilege EscalationSkillJect 1.0 (test)
ASR98
8
Information DisclosureSkillJect 1.0 (test)
ASR10,000
8
Efficiency AnalysisSkillJect
Tokens Used2,600
3
False Positive EvaluationSkillJect (Benign)
Ori FP Rate0
3
Malicious Skill DetectionSkillJect (Malicious)
Originality98
3
Showing 8 of 8 rows