skill-testing-framework
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local platform script using Node.js at
.agents/skills/platform/skill-lab.mjsto validate skill structure and behavior. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external skill files, which allows potentially malicious instructions in those files to influence the agent during validation. 1. Ingestion points: External skill files found at the user-provided path. 2. Boundary markers: None specified to separate analyzed content from the framework instructions. 3. Capability inventory: Executes a Node.js process and generates Markdown documentation based on processed data. 4. Sanitization: No explicit content filtering or sanitization of the input skill files is mentioned.
Audit Metadata