chrome-store-submit

Pass

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches the Pillow library and the @playwright/mcp package (provided by Microsoft) for image processing and browser control.
  • [COMMAND_EXECUTION]: Runs a local Python script (scripts/generate_store_images.py) to transform the extension's icon into required promotional images.
  • [COMMAND_EXECUTION]: Employs Playwright-based MCP tools to navigate the Google Chrome Web Store developer console and fill forms.
  • [PROMPT_INJECTION]: The skill ingests data from local manifest.json and README.md files to generate marketing copy, creating an indirect prompt injection surface.
  • Ingestion points: Local extension metadata and documentation files are parsed at the start of the workflow (SKILL.md Step 1).
  • Boundary markers: No specific delimiters are used to separate untrusted file content from system prompts during interpolation.
  • Capability inventory: The skill has access to shell execution for image generation and full browser automation (navigation, typing, and file upload).
  • Sanitization: Security is enforced by requiring the user to log in manually and review all AI-generated descriptions before they are entered into the store dashboard.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 27, 2026, 02:56 PM