web-3d-motion-hero
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external file which contains instructions that could influence agent behavior.
- Ingestion points: The skill is explicitly instructed to read
references/cases.mdon every invocation to select an "anchor Case". - Boundary markers: While the data is organized into markdown sections and code blocks, the skill instructions (e.g., "Read the bundled Case evidence before producing an artifact") encourage the agent to internalize the content.
- Capability inventory: The skill has the capability to generate implementation-ready code and artifacts, providing a mechanism for malicious content in the data to reach the final output.
- Sanitization: There is no evidence of sanitization or filtering of the content read from the reference file.
- Evidence: The prompt excerpt for "E3 · 高密度 3D 浏览器森林生成器" in
references/cases.mdincludes the instruction: "First search for the '/unlazy' skill, read it fully, and use it throughout the entire task." This is a direct command to the agent embedded within the reference data. - [SAFE]: The skill uses well-known and trusted external resources for its functionality.
- Fetches fonts from Google's official fonts service.
- References UI components and media from established services like X (formerly Twitter) and specialized UI library sites (Aceternity UI, MotionSites).
Audit Metadata