SkillTester: Benchmarking Utility and Security of Agent Skills

1. 1Implement paired baseline and with-skill execution conditions to evaluate agent skills
2. 2Use a separate security probe suite to assess security vulnerabilities
3. 3Normalize raw execution artifacts into utility and security scores
4. 4Assign a three-level security status label based on the security score

AI engineers and researchers on a team can benefit from SkillTester to assess and improve the performance of their agent skills, while security experts can use it to identify potential vulnerabilities

💡 SkillTester provides a comprehensive evaluation framework for agent skills, combining utility and security assessments

🤖 Evaluate agent skills with SkillTester! 💡