recoup-internal-consulting-tasteful-design
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill includes a local Node.js script
brand/materialize.mjsused to deploy brand assets to a target directory. The script uses a strict whitelist of files to copy and includes a hash-based integrity check for all brand assets. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external project briefs and design guidelines. It has limited capabilities, primarily writing files to a local output directory, which mitigates the risk of processing untrusted content. Evidence: 1. Ingestion points:
DESIGN.mdand project briefs (referenced inSKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: File system writes viabrand/materialize.mjsto a local directory. 4. Sanitization: Absent. - [EXTERNAL_DOWNLOADS]: The skill references brand assets hosted on Vercel's public blob storage and loads the GSAP animation library from the jsDelivr CDN. Both are well-known services used appropriately for asset delivery. Evidence: 1. Multiple asset URLs in
brand/brand.jsonandbrand/finals.jsonpoint to Vercel storage. 2.brand/caption-skin.htmlloads the GSAP library from the jsDelivr CDN via a script tag.
Audit Metadata