excalidraw-diagram

Pass

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute setup commands, such as uv sync and playwright install chromium, which are necessary to initialize the diagram rendering environment.\n- [EXTERNAL_DOWNLOADS]: The skill downloads essential components from well-known and established services, including the Playwright browser automation framework and the Excalidraw library via the esm.sh content delivery network.\n- [DYNAMIC_EXECUTION]: The render_excalidraw.py script employs Playwright to render diagrams by loading a local HTML file and executing JavaScript within a headless Chromium instance.\n- [INDIRECT_PROMPT_INJECTION]: The skill generates structured Excalidraw JSON from user-provided natural language descriptions. While this is the core functionality, it represents an attack surface for indirect prompt injection.\n
  • Ingestion points: User descriptions in the agent conversation, as defined in SKILL.md.\n
  • Boundary markers: None identified in the skill instructions to delimit untrusted input.\n
  • Capability inventory: File reading and writing (JSON and PNG) and execution of a local Python rendering script.\n
  • Sanitization: No input sanitization or validation mechanisms were observed.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 10, 2026, 11:27 AM
Security Audit — agent-trust-hub — excalidraw-diagram