public-ad-library-intelligence
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external third-party sources (Meta, Google, LinkedIn, TikTok, Reddit ad libraries) which could theoretically contain malicious instructions embedded in ad copy or metadata.
- Ingestion points: Data is retrieved via the
pull_public_ad_librarytool from providers like Adyntel and Apify (SKILL.md). - Boundary markers: The instructions include directives to "summarize patterns only" and "convert... into original creative concepts" while explicitly forbidding the reuse of competitor images, copy, or testimonials (SKILL.md).
- Capability inventory: The agent has the capability to fetch external data and generate creative briefs/concepts based on that data.
- Sanitization: No technical sanitization (e.g., escaping or schema validation of raw ad text) is explicitly described, relying instead on high-level instructions to the agent.
Audit Metadata