ad-copy
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill instructs the agent to omit warnings or caveats directly from the generated copy drafts ("Do not return a draft with a warning on it"). While this is intended to keep drafts clean for the user, it results in potential risks or assumptions being excluded from the final deliverable.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface. 1. Ingestion points: The agent fetches external websites and reads user-provided brand kits and existing copy as part of its core function. 2. Boundary markers: The skill provides explicit instructions in SKILL.md and references/agent-security.md to treat untrusted content as data rather than instructions. 3. Capability inventory: The skill can perform network requests, read local files, and write findings to .agents/product-context.md. 4. Sanitization: Sanitization for text output relies on the agent's adherence to safety instructions; HTML escaping is only specified for code generation in referenced security documentation.
Audit Metadata