funnel
Pass
Audited by Gen Agent Trust Hub on Apr 25, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions and reference files focus entirely on landing page architecture, marketing frameworks (such as AIDA, PAS, BAB, and 4Ps), and frontend performance best practices. No malicious code, prompt injection, or data exfiltration patterns were identified. The performance optimizations (e.g., preconnect, preload, fetchpriority) and modern CSS usage (e.g., View Transitions API, scroll-driven animations) follow established industry standards. Reference to well-known services like Vercel and Stripe is used neutrally to illustrate design patterns.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from other agents (Vision, Prose, Cast, Pixel) and user-provided briefs. This creates a surface for indirect prompt injection, but the skill lacks high-risk capabilities that would make such an injection dangerous.\n
- Ingestion points: Inputs from Vision (design direction), Prose (copy drafts), Cast (persona data), and user briefs in the
BRIEFphase.\n - Boundary markers: None identified in the handoff templates or core instructions.\n
- Capability inventory: No subprocess execution, dynamic code evaluation (exec/eval), file writing, or arbitrary network operations are present in the skill's own instructions or scripts.\n
- Sanitization: None specified for external content.
Audit Metadata