The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill interacts with api.browser-use.com and browser-use.com to provision cloud browsers and fetch API documentation. These resources belong to the official vendor infrastructure.
[DYNAMIC_EXECUTION]: The skill instructs the agent to use await import() with dynamic path concatenation and timestamp-based cache-busting to load and execute scripts stored in the ./.bcode/agent-workspace/ directory.
[INDIRECT_PROMPT_INJECTION]: As a tool designed for browser automation, this skill inherently processes untrusted data from web pages, creating a vulnerability surface where malicious content could attempt to influence the agent's behavior.
Ingestion points: External data retrieved from browser sessions (e.g., page titles, scraped DOM elements) via browser_execute and Runtime.evaluate calls.
Boundary markers: None are defined; the skill processes raw data from web targets.
Capability inventory: The toolset includes browser-level network operations, file writing to the workspace via the write tool, and arbitrary JavaScript execution within the browser context.
Sanitization: The skill does not provide specific instructions or mechanisms for sanitizing or escaping web-sourced content before it is interpolated into agent logic.

browser-execute