building-restaurant-site
Pass
Audited by Gen Agent Trust Hub on May 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a Node.js script (scripts/restaurant_site_data.js) to invoke Dineway CLI tools via
spawnSync. This is a functional requirement for uploading media to the CMS and is implemented securely using argument arrays to prevent shell injection. - [EXTERNAL_DOWNLOADS]: The helper script downloads restaurant images from external URLs found in enriched data files. It performs basic validation on MIME types and URL patterns to ensure only image assets are processed.
- [SAFE]: The skill processes untrusted restaurant data from 'places/*.json' files. Boundary markers are established via the 'Customer Visibility Gate' logic which mandates specific filtering. Its capability inventory includes file system writes and media uploads via the 'dineway' CLI. Sanitization is enforced through a 'Brand Voice & Copy Tone Brief' that ensures content is rewritten from a restaurant-owned perspective, preventing indirect prompt injection or exposure of raw external strings.
- [SAFE]: All identified dependencies and external tools are associated with the vendor 'dineway' or standard web development ecosystems like Astro and React.
Audit Metadata