visual-alignment-audit
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting data from external websites (live reference URLs). While this constitutes an attack surface for indirect prompt injection, the skill utilizes pre-defined JavaScript templates for specific property extraction (computed styles, layout dimensions), which significantly limits the potential for external content to manipulate agent behavior. The functionality is restricted to the skill's primary purpose of visual auditing.
- [COMMAND_EXECUTION]: The skill instructions include the use of
python3 -m http.server, which is a standard, safe method for serving local files during web development and does not involve executing untrusted shell commands. - [DYNAMIC_EXECUTION]: The agent is instructed to use
mcp__claude-in-chrome__javascript_toolto execute benign measurement scripts within a browser context. These scripts are provided as static templates within the skill to extract CSS and layout metadata, posing no risk of arbitrary code execution.
Audit Metadata