The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a framework for tools that process input from an AI agent, which represents a potential injection surface if the agent's context contains untrusted data.
Ingestion points: Tool arguments defined in netlify/functions/mcp.ts via the server.tool method.
Boundary markers: The skill advises using explicit instructions in tool descriptions to guide model behavior.
Capability inventory: Logic is developer-defined; example provided uses read-only operations (getItem).
Sanitization: Mandatory use of zod for strict schema validation of all inputs.
[EXTERNAL_DOWNLOADS]: Recommends installing standard, well-known libraries such as @modelcontextprotocol/sdk and zod from the official NPM registry.
[COMMAND_EXECUTION]: Provides instructions for using standard development tools like netlify-cli and the @modelcontextprotocol/inspector for local testing.

netlify-mcp-servers