The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is entirely built around the use of the agent-browser CLI tool. It provides a comprehensive map of commands and includes instructions for installing browser binaries via agent-browser install if they are missing from the environment.
[INDIRECT_PROMPT_INJECTION]: As a tool designed to interact with external web content, the skill possesses an inherent attack surface for indirect prompt injection. Malicious instructions placed on a website could potentially be interpreted by the agent during the open, get, or snapshot phases.
Ingestion points: External web page content accessed via navigation and inspection commands in SKILL.md.
Boundary markers: None provided in the command templates or instructions.
Capability inventory: The skill can execute JavaScript (eval), access session metadata (cookies, storage), and interact with the local filesystem (upload, screenshot, pdf).
Sanitization: No explicit sanitization or filtering logic is defined within the skill instructions.
[DYNAMIC_EXECUTION]: The agent-browser eval command allows for the execution of arbitrary JavaScript code strings within the browser session. While this is a standard feature for advanced browser automation, it represents a dynamic execution vector.

browse