The Agent Skills Directory

[External Resource Integration]: The skill references established libraries and deployment templates from reputable sources. Evidence: Documentation in deploy-cloudflare-workers.md and remote-http-scaffold.md recommends using the official @modelcontextprotocol/sdk, fastmcp, and Cloudflare templates. These are well-known resources within the development community and are appropriate for the skill's purpose.
[Data Ingestion and Processing]: The instructions facilitate the creation of tools that interact with external data sources, presenting a standard surface for handling untrusted data. Evidence: Scaffolds in remote-http-scaffold.md and resources-and-prompts.md describe tools that retrieve information from upstream APIs and the local filesystem.
[Ingestion Points]: Servers built using these patterns typically read from external SaaS APIs or local files.
[Boundary Markers]: The provided scaffolds rely on standard tool output formatting; the skill suggests using the 'Instructions' primitive to provide hints to the agent.
[Capability Inventory]: The resulting servers can perform network requests and filesystem access based on the logic implemented in their tool handlers.
[Sanitization]: The skill promotes using Zod for strict input schema validation to ensure data integrity.
[Secure Credential Handling]: The skill provides clear guidance on protecting sensitive information, following established best practices. Evidence: In auth.md and deploy-cloudflare-workers.md, the documentation explicitly advises against hardcoding API keys or tokens, recommending the use of environment variables, secrets management tools, or the OS keychain for local servers.

build-mcp-server