Agent Skills – Open Security Database
5 hours ago
- #Skill Definitions
- #AI Security
- #Risk Analysis
- The Skills Security Index provides security risk analysis for agentic AI skill definitions.
- AI agents rely on modular skills, making skill instructions a critical attack surface.
- The index helps security engineers and developers understand potential risks before deployment.
- Each entry in the index represents a unique skill from major platform registries like GitHub.
- A deep scan of the skill's identity, instructions, and associated code is performed.
- Analyses focus on instructional risk, such as bypassing guardrails or performing sensitive operations.
- Risk is ranked dynamically across three dimensions: Pass, Low, Medium, High, and Critical.
- Instructions are classified into buckets like Tools, Code Execution, Web Access, etc.
- Findings report deviations from security best practices, such as Prompt Injection or Credential Exposure.
- Permissions are evaluated based on whether they are justified by the skill's purpose.