visual-testing
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a systematic visual testing workflow using standard and well-documented MCP tools for browser automation and memory storage.\n- [INDIRECT_PROMPT_INJECTION]: The skill interacts with external web content by navigating to target URLs and capturing screenshots.\n
- Ingestion points: The skill navigates to target URLs and captures page content using the claude-in-chrome tool set (SKILL.md).\n
- Boundary markers: The workflow mandates a PLAN Phase to define targets and thresholds, providing context isolation (SKILL.md).\n
- Capability inventory: The skill uses navigate, resize_window, computer screenshot/zoom, and memory_store tools (SKILL.md).\n
- Sanitization: Content is processed via pixel-level diffing algorithms against stored baselines, which inherently ignores textual instructions (SKILL.md).\n- [DYNAMIC_EXECUTION]: The skill includes JavaScript logic for validation checks and comparison algorithms. These are documented as cognitive steps for the agent to follow when interacting with MCP tools and do not involve executing external scripts or dynamic runtime compilation (SKILL.md).
Audit Metadata