suede-visibility-grader
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data from public URLs and GitHub repositories, which creates a vulnerability surface for indirect prompt injection attacks where malicious instructions on a webpage could influence the agent's behavior.
- Ingestion points: In
SKILL.md, the agent is instructed to inspect live URLs, status codes, redirects, canonicals, robots.txt, sitemaps, page titles, rendered desktop/mobile viewports, visible headings (H1, H2), body copy, and metadata (Open Graph, Twitter cards, schema/JSON-LD). - Boundary markers: The instructions lack specific guidance on using delimiters or protective wrappers (e.g., XML tags or clear 'ignore embedded instructions' warnings) when processing fetched website content.
- Capability inventory: The skill performs complex multi-lane analysis (Findability, First-screen clarity, CTA pull, Proof and trust, AI readability, Design signal) and generates reports based on this external data.
- Sanitization: No sanitization, filtering, or escaping logic is defined for the content retrieved from external sources before it is processed by the LLM.
Audit Metadata