flowchart
Pass
Audited by Gen Agent Trust Hub on Jul 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell scripts (
scripts/tqandscripts/new-canvas) to interact with the local environment. These scripts facilitate the creation of diagram files and manage communication with the tldraw application's local HTTP server using standard tools likecurlandpython3. - [REMOTE_CODE_EXECUTION]: The skill generates JavaScript code based on user-provided process descriptions and executes it against the tldraw Editor API via a local endpoint (
/api/doc/:id/exec). This dynamic execution is the primary mechanism for programmatically building diagrams and is restricted to the local tldraw environment. - [PROMPT_INJECTION]: The skill ingests untrusted user input (prose descriptions of workflows) which is then used to generate logic for canvas manipulation. While this presents an indirect prompt injection surface, the capabilities are scoped to drawing operations within the local tldraw application.
- [COMMAND_EXECUTION]: Accesses a local configuration file (
server.json) within application support directories to retrieve the port and session token required for authenticated communication with the tldraw offline application.
Audit Metadata