The Agent Skills Directory

Dynamic Script Generation: The workflow involves generating a webmcp.init.js file and executing it within a browser environment using the agent-browser tool. This pattern is central to the skill's purpose of tool creation and testing.
Handling Untrusted Web Data: The skill interacts with external web pages to derive tool definitions and schemas. The instructions include a specific safety section advising the agent to treat all page-derived content as untrusted, which helps mitigate risks associated with indirect prompt injection from processed web data.
Command Line Interaction: The skill utilizes the agent-browser command-line utility to perform actions like opening URLs and invoking tools. This is standard behavior for the intended development and automation tasks, scoped to the browser environment.

webmcp-gen