conversational-ad-copy
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists entirely of instructional text and metadata. It does not contain any executable scripts, shell commands, or hidden code blocks.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, such as ad briefs, landing page content, and product feeds. While this creates an attack surface for indirect prompt injection, the skill includes robust mitigation strategies: a mandatory 'Grounding pass' to verify all claims against supplied sources and a 'Quotability gate' with 10 explicit checks to ensure output remains self-contained and neutral. No dangerous capabilities (like file writing or arbitrary network access) are exposed to this data.
- [SAFE]: All external references are directed toward well-known, legitimate advertising platforms (e.g., Google, Amazon, OpenAI) or internal helper skills within the same developer's namespace (@mbfinotti). There is no evidence of typosquatting or redirection to untrusted domains.
Audit Metadata