before-after-ad
Pass
Audited by Gen Agent Trust Hub on Jul 13, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized network operations were detected. The skill instructions demonstrate security best practices by defining clear stop conditions, production locks, and a human-in-the-loop requirement before handing off tasks to downstream media runners.
- [PROMPT_INJECTION]: The skill processes external data such as product facts and reference analysis, creating a potential surface for indirect prompt injection. This risk is effectively mitigated by specific instructions to lock parameters, adhere to claim boundaries, and stop execution if evidence is insufficient. The mandatory user approval step before production minimizes the risk of automated execution of malicious instructions. \n
- Ingestion points: The 'Ground The Brief' step in SKILL.md reads user-supplied product facts, claims, and references. \n
- Boundary markers: The skill uses logical 'Production Locks' and 'Claim Boundaries' as internal delimiters. \n
- Capability inventory: The skill facilitates hand-offs to image-batch-runner and video-batch-runner for asset generation. \n
- Sanitization: No explicit text sanitization is performed, but output is governed by strict rubrics and stop conditions.
Audit Metadata