hero-ascii-render
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements a rendering pipeline designed to process untrusted visual content from images, videos, and webcam streams, which represents a potential attack surface where adversarial visual patterns could influence the agent.
- Ingestion points: Untrusted visual data enters the system via the
hero-ascii__sourceimage/video elements and thenavigator.mediaDevices.getUserMediaAPI inSKILL.md. - Boundary markers: The implementation lacks explicit delimiters or instructions for the agent to treat visual input as strictly non-instructional, relying on the rendering pipeline to transform the data.
- Capability inventory: The skill uses the Canvas 2D API for data analysis and rendering (
getImageData,fillText) and includes functionality to request access to the user's camera. - Sanitization: Data is processed through downsampling and sRGB-to-linear luminance conversion, which reduces the data density but does not sanitize against semantically embedded visual prompts.
Audit Metadata