reference-design-contract
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill ingests untrusted data from external URLs, screenshots, and product notes to generate design documents. This creates a surface for indirect prompt injection where malicious instructions hidden in these references could influence the generated DESIGN.md or implementation files.\n
- Ingestion points: External URLs and image artifacts described in the Workflow (SKILL.md).\n
- Boundary markers: The skill instructions implement a 'Keep/Change/Do not copy' logic which provides some separation, but no technical delimiters are used for the external content.\n
- Capability inventory: The skill requires and uses the
file_writecapability to produce its output files (SKILL.md).\n - Sanitization: There is no evidence of content sanitization or validation for the ingested external data.
Audit Metadata