The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches protocol specifications and SDK documentation from official sources, including modelcontextprotocol.io and its associated subdomains. These downloads are essential for the development process and originate from well-known, trusted domains.
[COMMAND_EXECUTION]: The provided scripts (scripts/evaluation.py and scripts/connections.py) facilitate starting MCP servers using the stdio transport, which involves executing local commands and arguments provided by the user. This is a standard feature for development harnesses and is necessary for testing and verifying MCP server implementations.
[INDIRECT_PROMPT_INJECTION]: The skill documents the potential for indirect prompt injection when an agent interacts with external tools that return data. It proactively addresses this risk by providing best practices in references/mcp_best_practices.md, such as input validation with strict schemas (Pydantic/Zod), character limits, and output formatting to minimize the attack surface.
[SAFE]: The skill handles authentication for its evaluation harness via environment variables, adhering to secure development practices. No hardcoded credentials, malicious persistence, or obfuscation techniques were detected across the skill's source files.

mcp-builder