web-section-recreation-spec
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user inputs (such as page goals, content, and stack constraints) and structured 'Case' data from
references/cases.mdto generate implementation-ready code. - Ingestion points: The workflow relies on user-provided project details and the 'Prompt excerpt' fields within the bundled
references/cases.mdfile. - Boundary markers: The skill instructions mandate a 'reference contract' and 'Preserve/Replace/Avoid' logic to filter and constrain how evidence is applied, providing functional boundaries for the agent.
- Capability inventory: The skill is designed to generate code and component specifications; however, it does not explicitly list tool permissions in its frontmatter.
- Sanitization: The skill lacks explicit sanitization or delimiters (e.g., 'ignore embedded instructions') for the content extracted from the reference documentation.
- [EXTERNAL_DOWNLOADS]: The skill references various external assets and libraries to be included in the generated web artifacts.
- Trusted/Well-known sources: It fetches fonts from Google Fonts (
fonts.googleapis.com) and assets from AWS CloudFront (d8j0ntlcm91z4.cloudfront.net). - CDN Dependencies: The design cases reference standard web libraries including React, Tailwind CSS, and Framer Motion via
unpkg.comandcdn.tailwindcss.com. - Vendor Resources: The skill refers to its own platform ecosystem via
goodcase.aiandmotionsites.aifor case summaries and media previews.
Audit Metadata