eas-hosting
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides templates for creating API routes that ingest and process untrusted external data.
- Ingestion points: API handlers in
app/api/userequest.json()to parse request bodies andnew URL(request.url)to parse query parameters (seen inSKILL.md). - Boundary markers: The skill includes a 'Rules' section that explicitly instructs the agent to 'ALWAYS validate and sanitize user input', which acts as a design-time constraint.
- Capability inventory: The handlers exhibit network access via the
fetchAPI (e.g., proxing to OpenAI or Weather APIs). However, they are restricted by the Cloudflare Workers runtime, which prevents direct filesystem access as noted in the 'Missing/Limited APIs' section ofSKILL.md. - Sanitization: While the skill mandates sanitization in its rules, the code examples provide basic logical checks (e.g., checking for missing fields) rather than comprehensive sanitization logic.
Audit Metadata