design-review
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill captures screenshots of web pages, which can include internal development environments or private preview URLs. It then instructs the agent to upload these screenshots to an external Azure DevOps platform. While intended for design feedback, this creates a data exfiltration path for internal visual information.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted visual content from web pages to generate its design audit. This creates a surface where malicious elements on a page could attempt to influence the agent's report or trigger unintended actions in the subsequent development workflow.
- Ingestion points: Captured screenshots of external or internal web pages (SKILL.md).
- Boundary markers: None; the skill requires processing the full visual surface of the page.
- Capability inventory: Capturing screenshots of arbitrary URLs and posting reports/attachments to Azure DevOps (SKILL.md).
- Sanitization: None; the agent reviews the raw visual render of the web page.
Audit Metadata