The Agent Skills Directory

[SAFE]: The skill serves as a technical reference and implementation guide for WebMCP. It identifies and documents security considerations such as origin isolation, permissions policies, and the risk of prompt injection through tool outputs.\n- [EXTERNAL_DOWNLOADS]: The skill references various official and community-maintained packages (e.g., use-webmcp-tool, webmcp-types). It explicitly warns developers about package naming hazards and typosquatting, distinguishing between official Google/W3C resources and third-party community projects.\n- [INDIRECT_PROMPT_INJECTION]: The documentation includes specific sections on securing tools against injection attacks. It instructs developers to mark third-party content using the untrustedContentHint annotation and to sanitize all data before it is returned to an AI agent.\n- [COMMAND_EXECUTION]: The skill provides standard shell commands (curl, grep) as diagnostic tools for developers to verify their site's security headers, such as Permissions-Policy and Origin-Agent-Cluster.

webmcp