web-3d-motion-hero-by-197q59i
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the
references/cases.mdfile, which contains summaries and prompt snippets sourced from third-party design cases. If these external sources contain malicious instructions, they could potentially influence the agent's output.\n - Ingestion points:
references/cases.md(metadata and prompt excerpts).\n - Boundary markers: The content is organized into labeled evidence cards (E1-E8) with specific sections.\n
- Capability inventory: The agent is tasked with generating code, narratives, and using browser/media inspection tools.\n
- Sanitization: The
SKILL.mdmandates the creation of a "Preserve / Replace / Avoid" reference contract, which requires the agent to explicitly filter and analyze the source content before implementation.\n- [EXTERNAL_DOWNLOADS]: The skill requires the agent to access and inspect media files (MP4, JPG) from external domains to verify visual fidelity.\n - Evidence: Links to
media.goodcase.ai,goodcase.ai, andx.comare provided in the cases reference file.\n - Methodology: The agent is instructed to use available browser or media inspection tools to view these assets.
Audit Metadata