sketch-to-diagram
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implementation follows best practices for image processing by using local tools or standard vision APIs and explicitly warning users about potentially sensitive content in sketches.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied images which may contain text instructions designed to influence the agent's behavior during the recognition and structure phases.
- Ingestion points: User-provided file paths or base64-encoded images processed via the image-recognition MCP (SKILL.md, Phase 1).
- Boundary markers: Includes a requirement for human approval of the structured element list before proceeding to the redraw phase (SKILL.md, Phase 2, Step 6).
- Capability inventory: File system access for writing reports and diagram files; tool invocation through the configured MCPs.
- Sanitization: No specific filtering or sanitization of transcribed text from the vision API is defined.
Audit Metadata