The Agent Skills Directory

[SAFE]: The skill documents a tool developed by Vercel Labs, a trusted organization. All repository links and documentation URLs point to legitimate infrastructure.
[EXTERNAL_DOWNLOADS]: Installation instructions utilize standard package managers (npm) and official repositories. The skill also provides a method to download its own SKILL.md file directly from the official GitHub repository of Vercel Labs via curl.
[COMMAND_EXECUTION]: The skill extensively documents the use of the agent-browser CLI for browser automation. It includes examples for navigation, element interaction, and state management.
[REMOTE_CODE_EXECUTION]: Documentation is provided for the eval command, which allows executing JavaScript within the browser context. This is a core feature of the automation tool and is intended for programmatic interaction with web pages.
[PROMPT_INJECTION]: The skill is designed to process external web content, creating a surface for indirect prompt injection. However, it mitigates this by instructing agents to use structured JSON snapshots and deterministic element references ('refs') rather than relying on potentially spoofable visual or textual cues.
[CREDENTIALS_UNSAFE]: Examples and templates for authentication use generic placeholders (e.g., '') or trivial test strings (e.g., 'password123'), following standard documentation practices without exposing sensitive data.

agent-browser