The Agent Skills Directory

[SAFE]: The skill acts as an instructional framework for developing MCP servers. It does not contain malicious code, hidden instructions, or exfiltration patterns.
[REMOTE_CODE_EXECUTION]: The provided code snippets demonstrate how to make legitimate API calls to well-known services (e.g., GitHub API) using standard libraries like httpx and the Model Context Protocol SDK. These are presented as educational templates.
[CREDENTIALS_SAFE]: The skill explicitly instructs users to manage sensitive information like API keys and tokens through environment variables (e.g., os.environ['GITHUB_TOKEN']) and configuration files, rather than hardcoding them into the server logic.
[INDIRECT_PROMPT_INJECTION]: While the skill describes building tools that ingest external data, it proactively mitigates injection risks by mandating typed parameters (Zod/Pydantic), input validation at the boundary, and sanitization of agent-supplied parameters.
[COMMAND_EXECUTION]: The skill includes example configurations for running MCP servers via node and python. These are standard execution patterns for the protocol and are used to demonstrate local deployment.

MCP Builder