visual-validation
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes rendered interface content (screenshots, HTML/CSS) from external or project sources which are untrusted. This creates a surface for Indirect Prompt Injection where malicious instructions embedded in the UI could influence the agent.
- Ingestion points: Browser inspection data, screenshots, and automation output as described in the 'Evidence Boundaries' section.
- Boundary markers: None present. The skill does not provide instructions to ignore or delimit embedded text within the captured evidence.
- Capability inventory: The skill explicitly authorizes the agent to 'fix findings and recheck them,' which implies file system write access to modify source code.
- Sanitization: No sanitization of the rendered text or UI data is specified.
- [DYNAMIC_EXECUTION]: The 'Capture Ladder' instructions recommend the creation and execution of 'a small temporary script that opens the local app and captures screenshots.' This involves generating and running code at runtime to achieve automation goals.
Audit Metadata