The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references standard development tools and libraries for build optimization and runtime tasks. These include well-known packages like @next/bundle-analyzer, webpack-bundle-analyzer, sharp, puppeteer, lodash-es, date-fns, and @aws-sdk/client-*. These are standard dependencies in the Next.js ecosystem and are mentioned in an educational context for performance optimization.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface where an agent might be asked to deploy a user-provided project. It correctly identifies the risk that builds execute project dependencies and explicitly instructs the agent to inspect them before exposing secrets.
Ingestion points: Project dependencies and codebase provided by the user for deployment (references/detailed-guide.md).
Boundary markers: The skill includes multiple 'Sharp Edges' and 'Validation Checks' that serve as guardrails to prevent accidental exposure of secrets.
Capability inventory: Build processes execute arbitrary dependencies; server-side components can potentially leak secrets via serialized data.
Sanitization: The skill provides guidance on using proper environment variable prefixes (NEXT_PUBLIC_) and runtime checks to isolate sensitive data.

vercel-deployment