verify-ui
Pass
Audited by Gen Agent Trust Hub on Jul 24, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill contains instructions such as "Never tell the user 'refresh and check'" and "DO NOT tell user to check" if a change is broken. These instructions are evaluated as benign instructional guidance for a developer agent. The goal is to ensure the agent performs its own quality assurance and delivers a verified result to the user, rather than a malicious attempt to hide unauthorized activity or bypass safety protocols.
- [COMMAND_EXECUTION]: The skill utilizes browser automation tools (identified as
mcp__claude-in-chrome__*) to navigate to local development servers, capture screenshots, and execute JavaScript. The JavaScript examples provided (e.g.,window.getComputedStyle) are restricted to inspecting element styles for layout debugging. These actions are consistent with the skill's stated purpose of visual UI verification.
Audit Metadata