The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user or agent to install the @opentiny/webmcp-cli package globally via npm. This is the core utility for the skill's functionality and is provided by the skill author.
[COMMAND_EXECUTION]: The skill operates by executing various shell commands to control browser tabs and interact with page-specific tools. This includes managing browser state, opening URLs, and running automated tasks on supported domains like CSDN, Juejin, and Excalidraw.
[DYNAMIC_EXECUTION]: The page-agent-tool includes an executeJavascript action that permits the AI agent to run arbitrary JavaScript code within the context of the current browser page. This is a highly capable feature intended for advanced page automation.
[INDIRECT_PROMPT_INJECTION]: The skill retrieves and analyzes content from external websites (via browserState and searchTree) which constitutes an attack surface where malicious web content could influence the agent's behavior.
Ingestion points: Web content and ARIA trees retrieved from any website navigated to by the agent using the webmcp-cli tool (specifically in SKILL.md and page-agent-tool descriptions).
Boundary markers: The instructions do not define explicit delimiters or 'ignore' instructions for the content retrieved from external DOM/Aria trees.
Capability inventory: The agent can perform browser navigation, click elements, fill forms, scroll pages, and execute JavaScript (detailed in SKILL.md and domains/ guides).
Sanitization: There are no documented sanitization or filtering steps applied to the web content before it is processed by the AI agent.

webmcp-cli-skill