excalidraw-diagram
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user text to create visual diagrams and writes the output directly to files in the current working directory.
- Ingestion points: User-supplied text content describing concepts, relationships, or hierarchies (SKILL.md, Workflow step 1).
- Boundary markers: The output uses Markdown structure and
%%delimiters to wrap the Excalidraw JSON, but these are for plugin compatibility and do not protect against instructions embedded in the source text. - Capability inventory: The skill utilizes a file-writing tool to automatically persist generated content to the disk without mandatory user confirmation for each write.
- Sanitization: While the skill replaces double quotes and parentheses with alternative characters for layout purposes, it does not filter or sanitize the input for malicious instructions or payload content that could target the agent or downstream Markdown viewers.
Audit Metadata