before-after-ad
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external information to generate prompts for downstream media generation tools. 1. Ingestion points: The skill processes user-supplied product facts, approved proof, reference analysis, and videos as defined in the 'Ground The Brief' section of SKILL.md. 2. Boundary markers: It implements 'claim_boundary', 'Production Locks', and 'continuity_locks' to constrain agent output. 3. Capability inventory: The skill is capable of handing off generated prompts to 'image-batch-runner' and 'video-batch-runner' tools. 4. Sanitization: A mandatory 'Review Checklist' and 'Stop Conditions' in 'references/shot-script-rubric.md' prevent the creation of unsupported claims, medical implications, or deceptive comparisons. Safety is further supported by a required user approval step before media production begins, although the 'references/downstream-handoff.md' file optimizes for speed by bypassing a second confirmation during the final handoff.
Audit Metadata