prototype
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user prompts to generate logic modules and UI variants. These inputs are interpolated directly into executable files (HTML script blocks and React components) without specified sanitization or validation. This creates a surface where a malicious prompt could inject scripts that execute when the prototype is opened or shared.
- Ingestion points: User prompts defining the prototype's logic or UI (SKILL.md).
- Boundary markers: Instructions lack requirements for delimiters or warnings to isolate user-provided data from generated code.
- Capability inventory: File system writes for HTML, JS, and TSX files; instruction to use task runners for execution (SKILL.md, UI.md).
- Sanitization: No evidence of escaping or filtering logic/UI descriptions before generation.
- [COMMAND_EXECUTION]: The skill instructions direct the creation of prototypes that are executable via common project tools such as pnpm, python, and bun. This encourages a workflow where automatically generated code is executed on the developer's machine with standard privileges.
Audit Metadata