drawio-diagram
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches documentation, sample diagrams, and templates from JGraph's official GitHub repositories (e.g.,
jgraph/drawio-mcp). These references target a well-known service and are verified to be from the official organization. - [COMMAND_EXECUTION]: Executes the local
drawiodesktop CLI to perform Mermaid-to-XML conversion, apply ELK layouts, and export diagrams to image formats. It also runs local Python scripts for XML and layout validation to ensure artifact quality. - [DYNAMIC_EXECUTION]: Provides platform-specific command templates (PowerShell, Node.js) for the agent to discover the Draw.io installation path and handle browser URL opening on Windows and WSL2 systems. These are standard workarounds for platform-specific limitations and installation variants.
- [INDIRECT_PROMPT_INJECTION]: As the skill processes user-provided diagram content and Mermaid source, it maintains an attack surface for indirect prompt injection. 1. Ingestion: Reads
.drawioXML and Mermaid files. 2. Boundaries: Uses markdown blocks for inputs. 3. Capability inventory: Includes shell execution and file writes. 4. Sanitization: Employs structural validation using standard XML parsers and specialized validation scripts to mitigate these risks.
Audit Metadata