The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation references the installation of developer utilities including portless and agent-browser from the npm registry. It also provides a workflow for utilizing Vercel Sandbox, a trusted environment for executing automation scripts, which involves fetching system dependencies and tool binaries from official repositories.
[COMMAND_EXECUTION]: The skill utilizes the agent-browser CLI to perform web automation tasks such as navigating to URLs, interacting with page elements, and executing JavaScript within the browser context. These actions are governed by a built-in safety hook (agent-browser-safety) that intercepts commands to ensure they comply with predefined security policies and user-configured limits.
[DATA_EXFILTRATION]: The skill includes commands for reading the system clipboard and capturing network traffic (HAR files). To mitigate the risk of data exposure, the skill provides strict rules for securing session states, recommending the use of an encrypted auth vault and enforcing restrictive filesystem permissions (chmod 600) for all sensitive state files.
[SAFE]: This tool primarily serves as a security layer for browser automation. It incorporates several safety features: it blocks access to internal network ranges and sensitive OAuth endpoints, honors robots.txt directives by default, and implements a nonce-based content boundary system to help the agent distinguish between its own instructions and untrusted data retrieved from external websites.

browser-tools