doc-image-agent

Pass

Audited by Gen Agent Trust Hub on Apr 17, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill relies on well-known and trusted external packages, specifically the requests library for Python and the Playwright MCP package for Node.js.
  • [EXTERNAL_DOWNLOADS]: Instructions are provided to download and install official browser runtimes via the playwright install command.
  • [COMMAND_EXECUTION]: The workflow involves executing a bundled Python script (scripts/generate_image.py) and browser automation tools to capture and process images.
  • [DATA_EXFILTRATION]: The skill performs legitimate network communication with the OpenRouter API to fulfill image generation requests based on user-provided prompts.
  • [CREDENTIALS_UNSAFE]: Security best practices are followed for credential management; API keys and site logins are accessed via environment variables, with explicit instructions to prevent exposure or hardcoding.
  • [SAFE]: The skill ingests data from Markdown files to drive its automation. While this represents a surface for indirect instruction, the risk is mitigated by the specific marker formats and the context of its primary document-processing purpose. \n
  • Ingestion points: Markdown files located in cases/{article-id}.md. \n
  • Boundary markers: Not explicitly used for the input document content. \n
  • Capability inventory: Includes browser automation (mcp_playwright_browser_*), shell command execution (via the generation script), and file system writes. \n
  • Sanitization: Not explicitly implemented for the ingested Markdown markers.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 17, 2026, 12:34 PM