platform-environment-skill-audit
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill functions as a security audit template, instructing agents to identify and remove sensitive information (credentials, tokens, personal paths) from other skills before publication.
- [INDIRECT_PROMPT_INJECTION]: The skill's purpose is to audit potentially untrusted external files and command instructions, creating a surface for indirect prompt injection. 1. Ingestion points: The skill body, linked files, and proposed commands (SKILL.md). 2. Boundary markers: The instructions explicitly direct the agent to 'Inspect' and 'Classify' content, treating it as data to be analyzed. 3. Capability inventory: Diagnostic operations such as 'resolve symlinks', 'verify every advertised path', and 'Confirm package... versions live' (SKILL.md). 4. Sanitization: The audit process requires the removal of secrets rather than just masking them and the categorization of facts into 'portable' or 'sensitive' scopes.
- [COMMAND_EXECUTION]: The audit requires verifying the existence of file paths and checking versions of installed drivers and toolkits on the local system, which are necessary diagnostic actions for its intended purpose.
Audit Metadata