The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and act upon data from external sources, including development server error overlays, terminal build outputs, and documentation pages. This represents a potential surface for indirect prompt injection where data content could influence the agent's decisions.
Ingestion points: Processes output from next dev, next build, and local documentation files in node_modules/next/dist/docs/.
Capability inventory: The skill can execute shell commands (npx, grep, find) and modify the local filesystem via codemods.
Boundary markers: The instructions do not explicitly define boundary markers for untrusted data processed from the dev overlay or build logs.
Sanitization: No specific sanitization or validation logic is described for the dynamic data ingested during the adoption loop.
External Downloads and Tool Installation: The skill installs additional components and tools from remote sources to facilitate the migration.
Evidence: The skill executes npx skills add https://github.com/vercel/next.js/tree/canary/skills/next-dev-loop and various npx @next/codemod commands. These resources originate from official Vercel repositories, which are recognized as trusted vendor sources.
Command Execution: The skill relies on executing local development tools to verify code changes and build the application.
Evidence: It invokes next dev and next build to validate the migration progress and identify blocking routes. This is standard behavior for an agent assisting with framework-level migrations.

next-cache-components-adoption