dym-diagram-design-import-extractors-shell
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for running local Python scripts via the terminal to process and extract data from diagram files.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted diagram source files and converts them into an intermediate representation (IR) that the agent later consumes. 1. Ingestion points: Files with extensions .drawio, .xml, .mmd, and Markdown files containing Mermaid blocks. 2. Boundary markers: The skill generates a specific 'digest' or 'IR JSON' format; the documentation explicitly notes that external label content is untrusted. 3. Capability inventory: The agent reads the extracted IR to perform redesigns, simplifications, or architectural analysis. 4. Sanitization: The skill documentation states that the scripts (specifically mermaid_extract.py) discard executable elements like click targets and render labels as inert text to mitigate risks.
Audit Metadata