tanstack-markdown-custom-extensions
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill documents an architecture for processing untrusted markdown data through custom extensions. This represents an indirect prompt injection surface where external data could potentially contain instructions intended to influence the agent. The documentation addresses this risk by providing guidelines on maintaining trust boundaries and using portable component nodes instead of raw HTML for untrusted inputs.
- Ingestion points: Markdown source strings processed by
parseMarkdownandrenderHtml(SKILL.md). - Boundary markers: The documentation explicitly identifies "Rich output versus untrusted-content safety" as a design tension and recommends the use of
ComponentNodeto maintain safety boundaries. - Capability inventory: The skill demonstrates capabilities for AST transformation and HTML generation via hooks.
- Sanitization: Includes specific guidance on treating HTML hooks as explicit trusted boundaries and suggests using framework-specific components to manage rich output safely.
Audit Metadata