representing-systems
Pass
Audited by Gen Agent Trust Hub on Aug 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The generated HTML documentation records fetch the Mermaid visualization library from JSDelivr, a well-known and reputable content delivery network.
- [REMOTE_CODE_EXECUTION]: The HTML record template initializes the Mermaid library using
securityLevel: "loose". This configuration allows the library to render HTML labels within diagrams, bypassing default sanitization and creating a potential path for script execution if malicious content is included in the system data. - [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by processing external system metadata into documentation records that execute JavaScript. 1. Ingestion points: System labels and component metadata in
templates/registry/records/system/1.html. 2. Boundary markers: Absent. 3. Capability inventory: Local file creation and browser-based code execution via the Mermaid renderer. 4. Sanitization: Absent; the implementation explicitly relaxes security constraints in the visualization layer.
Audit Metadata