drawio-skill

Pass

Audited by Gen Agent Trust Hub on May 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes the draw.io desktop application CLI to perform diagram exports and metadata embedding. It also executes bundled Python scripts (repair_png.py and encode_drawio_url.py) to handle binary file corrections and string encoding. Standard system utilities like mkdir and platform-specific open commands are used for file management.- [EXTERNAL_DOWNLOADS]: Provides guidance for downloading the required draw.io application from official and trusted sources, including Homebrew and the official jgraph repository on GitHub.- [PROMPT_INJECTION]: An indirect prompt injection surface is present in the style extraction workflow which processes untrusted diagram data.
  • Ingestion points: The skill parses content from user-provided .drawio XML files and various image formats (PNG, JPG) in references/style-extraction.md.
  • Boundary markers: The instructions do not define delimiters or markers to isolate untrusted label data from the agent's context.
  • Capability inventory: The skill executes shell commands and writes to the local filesystem.
  • Sanitization: Extraction logic uses regex for semantic role mapping but does not implement filtering or sanitization for potential malicious instructions within diagram labels.
Audit Metadata
Risk Level
SAFE
Analyzed
May 17, 2026, 02:56 PM