tune-mobile-client-performance
Pass
Audited by Gen Agent Trust Hub on Aug 2, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill defines a process for ingesting external evidence such as profiler traces and network captures, which constitutes an attack surface for indirect prompt injection. \n- [SAFE]: The 'Trust boundary' section provides explicit instructions to treat all external evidence as non-authoritative and forbids the execution of embedded commands or the disclosure of secrets. \n- [SAFE]: The instructions mandate the sanitization of private content and identifiers, ensuring that diagnostic results do not inadvertently leak sensitive user data. \n- [SAFE]: The skill includes evaluation test cases that verify the agent's ability to refuse malicious instructions embedded in diagnostic notes, such as requests to print authentication tokens or upload traces to untrusted endpoints.
Audit Metadata