honest-review
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from automated analysis tools that may ingest attacker-controlled content from generated media.
- Ingestion points: Data entering via the 'gooseworks:media_analyze' tool, including transcripts and visible text from generated assets.
- Boundary markers: Instructions define specific reporting prefixes such as 'the automatic read shows' to distinguish tool outputs from agent observations.
- Capability inventory: The skill utilizes analysis reading capabilities and references a 'save-learning' skill for recording brand rules.
- Sanitization: There is no programmatic sanitization mentioned for ingested text; the process relies on human confirmation ('please confirm') as a mandatory check before findings are finalized.
Audit Metadata