scoring-agent-skills
Pass
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill directs the agent to run validation tools to assess the quality of other skills. It establishes safety boundaries by instructing the agent not to execute untrusted code unless it is sandboxed and explicitly authorized by the user.
- [PROMPT_INJECTION]: As a tool designed to inspect and process third-party repository content, the skill is subject to indirect prompt injection where malicious instructions in the analyzed files could attempt to manipulate the scoring process. * Ingestion points: SKILL.md files and directly linked resources from the repositories being evaluated. * Boundary markers: The instructions include logic for identifying untrusted sources but do not provide explicit delimiters for isolating external data during processing. * Capability inventory: File system read access and shell command execution for running validators. * Sanitization: The instructions explicitly require sandboxing and user authorization for any code execution involving untrusted sources.
Audit Metadata