The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides code templates for API routes and middleware that process untrusted data from external sources, creating a potential surface for indirect prompt injection.
Ingestion points: The template src/app/api/users/route.ts ingests data from request.nextUrl.searchParams and request.json() bodies. The src/middleware.ts template ingests data from request.cookies.
Boundary markers: No specific delimiters or "ignore instructions" warnings are included in the prompt templates or the generated code to mitigate injection risks in processed data.
Capability inventory: The skill is configured with Read, Write, Edit, and Grep tools, which allow the agent to modify project files based on interpreted content.
Sanitization: The provided example code (Step 4) demonstrates basic type casting for search parameters but lacks comprehensive input validation or sanitization before passing the request body to the database layer.

vercel-reference-architecture