grill-with-visuals
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill employs local Node.js scripts (
glossary.mjs,shoot.mjs,chartcheck.mjs) to automate visualization tasks. These scripts perform standard file operations and environment checks within the user's local workspace, such as using a headless browser to verify the layout of generated pages.\n- [EXTERNAL_DOWNLOADS]: Remote references are limited to trusted and well-known sources. Thechartcheck.mjsscript verifies the presence of TLA+ tools from their official GitHub repository, and CSS/JS assets are loaded from standard CDNs (Google Fonts, jsDelivr) for rendering visual components.\n- [DYNAMIC_EXECUTION]: The skill dynamically loads the Mermaid diagramming library and ELK layout engine from a public CDN withingrill.js. This is a standard implementation used solely for rendering charts and diagrams in the generated HTML reports.\n- [DATA_EXFILTRATION]: No exfiltration patterns were detected. The skill instructions specifically require that all generated visual rounds stay as local files and are never published as artifacts, ensuring data remains within the local environment.\n- [INDIRECT_PROMPT_INJECTION]: While the skill ingests content from other agent roles to populate templates, it incorporates security best practices such as HTML entity escaping inglossary.mjsand the use ofdata-no-glossaryboundary markers to prevent the misinterpretation of embedded data.
Audit Metadata