ad-creative
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external landing page URLs provided by the user to perform a message-match assessment, which creates a surface for instructions embedded in web content to influence agent behavior.
- Ingestion points: Untrusted external content from user-provided landing page URLs.
- Boundary markers: The instructions do not specify any delimiters or safety warnings to ensure the agent ignores instructions found within the fetched web data.
- Capability inventory: The skill reads local files in the
~/.claude-marketing/directory (profiles, guidelines, and SOPs) and references several other tools for metadata and compliance checking. - Sanitization: No evidence of input sanitization or filtering of the external content before it is processed by the AI.
Audit Metadata