marketing-expert
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes an architecture for 'Intent Prediction' and 'Dynamic Pathing' where external, untrusted data (mouse movements, scroll depth, click patterns, and browsing history) is used to determine agent orchestration and UI adjustments.
- Ingestion points: User interaction data, browsing history, and 'Zero-Party Data' described in
references/ai-personalization.md. - Boundary markers: No explicit boundary markers or 'ignore' instructions are provided to separate user-derived intent signals from system instructions.
- Capability inventory: The skill uses
activate_skillto chain operations (ui-ux-pro,seo-pro) and references scripts (scripts/calculate-clv.py,scripts/lint-cro-check.ts) to perform analysis. - Sanitization: There is no mention of sanitizing or validating the behavioral data before it influences the agent's sequential activation logic.
Audit Metadata