The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute browser automation commands via a CLI tool and shell. This functionality is intended for programmatic web interaction.
[REMOTE_CODE_EXECUTION]: The skill provides an eval command that allows the execution of JavaScript within the browser environment. This is a primary feature for complex data extraction and interaction, and the documentation provides methods to use it securely via stdin or Base64 encoding.
[EXTERNAL_DOWNLOADS]: The skill references the use of npx and appium for package execution and mobile device automation. These are standard development tools and are used as intended for the skill's functionality.
[PROMPT_INJECTION]: The skill acknowledges the risk of indirect prompt injection from web content. It provides a specific mitigation feature called --content-boundaries that wraps tool output in nonces to help the agent differentiate between its instructions and untrusted page content. The ingestion points include the open, snapshot, and get commands, while capabilities include browser-based eval and file writing (screenshots).

agent-browser