web-design-reviewer
Pass
Audited by Gen Agent Trust Hub on Mar 26, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted web content from local or remote URLs to identify UI issues, which creates a surface for indirect prompt injection.
- Ingestion points: Website DOM structure and visual snapshots are retrieved using browser automation tools (SKILL.md, Step 2.1).
- Boundary markers: No specific instructions or delimiters are used to prevent the agent from following instructions potentially embedded in the reviewed web content.
- Capability inventory: The agent has permission to search the codebase and read/write source files to apply design fixes (SKILL.md, Step 3.2, 3.3).
- Sanitization: No validation or sanitization of the retrieved website content is performed before processing.
- [EXTERNAL_DOWNLOADS]: The skill documentation provides a configuration example for the @playwright/mcp tool. This is a well-known package provided by Microsoft and retrieved from the official NPM registry.
- [COMMAND_EXECUTION]: The skill involves searching the local filesystem and modifying source code files in the workspace to resolve design issues, using tool calls for searching and editing code.
Audit Metadata