industry-standard-reviewer
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [Prompt Injection] (SAFE): The metadata and descriptions do not contain instructions meant to override agent behavior or bypass safety filters.
- [Data Exposure & Exfiltration] (SAFE): The skill does not access sensitive file paths or initiate network requests.
- [Unverifiable Dependencies & RCE] (SAFE): No external Python or Node.js packages are required, and there are no patterns for downloading or executing remote code.
- [Indirect Prompt Injection] (LOW): 1. Ingestion points: The
submission_diffinput accepts untrusted user-supplied code. 2. Boundary markers: Absent in the YAML schema. 3. Capability inventory: The skill possesses no capabilities to interact with the system, network, or files. 4. Sanitization: Not specified. Risk: While an attacker could put instructions in a diff, the skill has no dangerous tools for those instructions to exploit.
Audit Metadata