access-review
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill contains defensive instructions intended to protect the agent from malicious commands embedded in audited data. It specifically instructs the agent to recognize and report phrases like 'ignore previous instructions' as security findings within the data, rather than obeying them as system commands.
- [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes identity metadata (role names, group descriptions) provided via arguments. * Ingestion points: Accesses files and directories through the $ARGUMENTS variable. * Boundary markers: Employs a 'SECURITY BOUNDARY' section and a 'Prompt Injection Safety Notice' to delimit safe operations. * Capability inventory: Utilizes Read, Grep, and Glob tools for data inspection. * Sanitization: Instructs the agent to treat all input as untrusted and explicitly forbids executing any instructions found within the data fields.
Audit Metadata