prototype
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill defines a constrained environment for code generation. Instructions in LOGIC.md specify the use of self-contained HTML files with no external dependencies, bundlers, or servers. UI.md specifies using existing project components and gating prototype features in production builds using environment variables.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input through user prompts and existing codebase context. 1. Ingestion points: User-provided descriptions of logic or UI requirements and existing project files (SKILL.md). 2. Boundary markers: None; the agent relies on its inherent training to distinguish instructions from data. 3. Capability inventory: Potential subprocess calls for task runners (pnpm, python, bun), file system writes for new routes and demo files, and Git operations (SKILL.md, LOGIC.md, UI.md). 4. Sanitization: None; input is interpreted as design requirements.
Audit Metadata