cursor-animated-sites
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill generates HTML files based on a structured data model, which defines how the agent should interpret and display steps. This architecture introduces a potential surface for indirect prompt injection if the content being visualized includes malicious instructions.
- Ingestion points: The
STEPSdata structure inSKILL.mdwhich includes narration and title fields used to build the walkthrough. - Boundary markers: The generation instructions do not specify the use of clear delimiters or instructions to ignore embedded commands within the visualization data.
- Capability inventory: The skill instructions include writing to the local file system (specifically
llmwiki/html/andwiki/paths) and invoking apython3command for HTML parsing and validation as described inSKILL.md. - Sanitization: The provided logic does not include explicit routines for sanitizing or escaping the data before it is rendered into the HTML output.
Audit Metadata