sequence-diagram

Pass

Audited by Gen Agent Trust Hub on Jul 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes local shell scripts (scripts/tq and scripts/new-canvas) to automate diagram creation and communicate with the tldraw offline application. These scripts perform operations such as copying templates, opening the application, and sending commands to the local server.
  • [DATA_EXFILTRATION]: The skill reads the local server.json configuration file created by the tldraw offline app to obtain the port and authentication token necessary for API access. This is a standard procedure for the bridge and is confined to the local machine.
  • [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it processes user-supplied text or DSL to create diagram elements.
  • Ingestion points: User request prose or DSL model in SKILL.md.
  • Boundary markers: Not present; the agent is instructed to parse input directly.
  • Capability inventory: Local file creation, shell script execution, and JavaScript execution within the tldraw application via the bridge.
  • Sanitization: The skill parses input into a structured model, but the shell scripts do not implement explicit escaping for variables interpolated into JavaScript strings (e.g., in scripts/new-canvas), which could allow for script injection within the tldraw app if malicious file paths are used.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 26, 2026, 05:42 PM
Security Audit — agent-trust-hub — sequence-diagram