design-qa
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection Surface: The skill is designed to ingest and analyze external data from web URLs and design tools. This creates a potential surface for indirect prompt injection, where content within the designs or rendered pages could attempt to influence the agent's behavior. (File: SKILL.md)
- Ingestion points: External URLs and visual design targets.
- Boundary markers: None identified.
- Capability inventory: Browser navigation, screenshot capture, and local command execution.
- Sanitization: Not explicitly defined for ingested content.
- Command Execution: The skill references running 'npm run check:runtime', which involves local command execution within the project environment. This is a standard part of the development workflow but represents a capability that should be monitored. (File: SKILL.md)
- Sensitive Context Interaction: The skill involves accessing local addresses such as 127.0.0.1 to capture screenshots for visual comparison. This is a core function for prototype QA but involves interaction with the local network environment. (File: SKILL.md)
- Potential Prompt Injection Pattern: The use of the header 'Critical Overrides' matches patterns typically associated with attempts to bypass or override instructions. In this context, it appears to be a reference to an external document, but the naming convention is a point of interest for review. (File: SKILL.md)
Audit Metadata