drawio-gcp

Pass

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the installation of the drawio-ai-kit package from the official vendor repository on GitHub (github:sparklabx/drawio-ai-kit).\n- [COMMAND_EXECUTION]: Uses the drawio-ai CLI to search for icons, manage workflows, and render diagram files into PNG format.\n- [DYNAMIC_EXECUTION]: Generates a JavaScript script (build.mjs) from templates and executes it locally using Node.js to programmatically construct the diagram layout.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided diagram requests by interpolating them into a subagent prompt, which could be used to pass hidden instructions.\n
  • Ingestion points: User's architectural request and clarifications are interpolated into the subagent text block in SKILL.md.\n
  • Boundary markers: The input is demarcated using <user's request + clarifications, verbatim> placeholders.\n
  • Capability inventory: The skill is capable of shell command execution, Node.js script execution, and file system write operations.\n
  • Sanitization: There is no explicit evidence of input sanitization or filtering before the data is passed to the subagent.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 10, 2026, 01:53 PM
Security Audit — agent-trust-hub — drawio-gcp