The Agent Skills Directory

[COMMAND_EXECUTION]: The skill scripts utilize Node.js child_process functions like spawn and execFileSync to interact with the browse CLI tool. This execution is required to connect to the Chrome DevTools Protocol (CDP) and perform background sampling of the browser's state.
[EXTERNAL_DOWNLOADS]: The skill requires the browse CLI tool to be installed globally via npm. This tool is provided by Browserbase, a well-known service in the browser automation space, and is the primary mechanism for the skill's functionality.
[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection because it is specifically designed to ingest and summarize external web data (DOM content, console logs, and network traffic) for an AI agent's consumption. While the skill's scripts are safe, any agent processing the resulting traces should treat the captured data as untrusted external input.
[SAFE]: The skill follows security best practices by using environment variables for sensitive API keys (BROWSERBASE_API_KEY) and providing clean, documented scripts that focus strictly on the intended observability use case.

browser-trace