The Agent Skills Directory

[SAFE]: The skill is a developer utility designed to facilitate the adoption of the WebMCP standard. It uses standard browser APIs and provides a local compatibility script for production use.
[PROMPT_INJECTION]: The skill mentions prompt injection patterns (e.g., 'SYSTEM: ignore previous instructions') exclusively within the context of security auditing and documentation. It explicitly instructs the agent to detect these as failure states in third-party integrations rather than executing them.
[INDIRECT_PROMPT_INJECTION]: The skill identifies the attack surface for indirect prompt injection in tool descriptions and outputs. It provides robust remediation guidance, including the use of boundary markers ('Write descriptions that describe — never instruct') and sanitization of user-generated content.
[DYNAMIC_EXECUTION]: The use of execution callbacks within the registerTool examples is a standard feature of the WebMCP API for connecting agents to page functionality. The implementation in the provided scripts is restricted to local scope and intended for developer testing.
[EXTERNAL_DOWNLOADS]: The skill references documentation and standards from trusted organizations, including the W3C (webmachinelearning.github.io) and Google (developer.chrome.com). These references are used for technical context and do not involve untrusted remote code execution.

webmcp-integration