frontend-visual-verify
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes component source code and navigates to browser-rendered pages, which constitutes a surface for potential indirect prompt injection.\n
- Ingestion points: Input arguments (Storybook URLs, file paths), component source code (JSX/TSX), and browser accessibility snapshots.\n
- Boundary markers: The skill utilizes a two-phase workflow (Analysis then Execution) with explicit user checkpoints via AskUserQuestion.\n
- Capability inventory: Access to tool-restricted file reading, writing to specific subdirectories, and browser navigation via Playwright MCP.\n
- Sanitization: Implements a 'localhost-only' default URL policy and requires manual approval before proceeding with verification steps or finalizing report writing.\n- [COMMAND_EXECUTION]: The skill performs legitimate file management and indexing tasks using scoped Bash tools and local plugin scripts.\n
- Evidence: The skill calls Bash tools for directory creation (mkdir) and listing (ls) within the permitted $JAAN_OUTPUTS_DIR. It also sources utility scripts from the plugin root for ID generation and index maintenance.
Audit Metadata