Malicious Skill Detection on 100-skill dataset (OpenClaw)
98TPRSkillScanner (LLM)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SkillScanner (LLM)variant=LLM-judgment2026.06 | 98 | 90 | 54 | |
| RSA2026.06 | 88 | 8 | 90 | |
| SkillScannervariant=default2026.06 | 70 | 12 | 77 | |
| Aguara2026.06 | 64 | 14 | 75 | |
| Semgrep (custom)ruleset=custom2026.06 | 34 | 0 | 67 | |
| ClawScan2026.06 | 2 | 6 | 48 | |
| Semgrep (generic)ruleset=generic2026.06 | 0 | 2 | 49 |