The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/setup.sh script performs a global installation of the agent-browser package from the NPM registry and subsequently downloads the Chromium browser binary.- [COMMAND_EXECUTION]: The skill operates by executing various CLI commands through the agent-browser tool to interact with the web, including open, click, fill, and snapshot.- [DYNAMIC_EXECUTION]: The eval command documented in references/commands.md allows the agent to execute arbitrary JavaScript code within the browser's execution environment.- [INDIRECT_PROMPT_INJECTION]: The skill represents a vulnerability surface for indirect prompt injection due to its core function of processing untrusted web data.
Ingestion points: External website content is ingested into the agent context via agent-browser open, agent-browser snapshot, and agent-browser get text (found in SKILL.md and references/commands.md).
Boundary markers: The documentation in SKILL.md suggests setting the AGENT_BROWSER_CONTENT_BOUNDARIES=1 environment variable to wrap tool output for safety.
Capability inventory: The skill possesses significant capabilities including element interaction (click, fill), arbitrary code execution (eval), file system writes (screenshot, pdf, state save), and credential management (auth save).
Sanitization: The skill provides optional configuration for domain allowlisting (AGENT_BROWSER_ALLOWED_DOMAINS) and action policies to restrict agent behavior.

agent-browser