The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches documentation from claude.com to stay updated with platform-specific connector constraints. It also provides instructions to bootstrap projects using official Cloudflare Workers templates and standard packages from NPM (@modelcontextprotocol/sdk, express, zod) and PyPI (fastmcp). All identified external references target trusted organizations or well-known technology services.
[INDIRECT_PROMPT_INJECTION]: The skill guides the construction of MCP servers that ingest natural language inputs through tool parameters (e.g., search queries in references/remote-http-scaffold.md or action intents). This creates a potential surface for indirect prompt injection if the processed data contains malicious instructions. However, the skill actively encourages security best practices to mitigate this, such as using strict Zod schemas for input validation and writing unambiguous tool descriptions to guide model behavior. The capability inventory for the generated servers includes network access and file system interaction, which are aligned with the skill's primary purpose of building API and system integrations.
[SAFE]: The skill demonstrates a strong security posture by including guidelines for secure authentication (OAuth 2.0, CIMD, DCR), recommending OS keychains over plaintext storage for local credentials, and advising on input sanitization through schema definitions. No indicators of obfuscation, privilege escalation, or malicious persistence were found.

build-mcp-server