The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the agent-browser CLI to control Chromium-based browsers via the Chrome DevTools Protocol (CDP). This allows for programmatic interaction with web pages and desktop applications like Slack or Discord.
[EXTERNAL_DOWNLOADS]: The SKILL.md file includes instructions to run agent-browser skills get <name>, which retrieves skill definitions and templates from a remote repository. The project synchronizes content from a repository maintained by a well-known and trusted technology organization.
[REMOTE_CODE_EXECUTION]: The skill documents the agent-browser eval command, which allows the execution of arbitrary JavaScript within the context of the automated browser page. This capability is used for complex interactions that cannot be achieved through standard CLI commands.
[CREDENTIALS_UNSAFE]: The documentation addresses the handling of sensitive session tokens and cookies. It provides security best practices, such as using environment variables for credentials, utilizing an internal auth vault, and enabling encryption for state files stored at rest via the AGENT_BROWSER_ENCRYPTION_KEY environment variable.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from web pages that could contain malicious instructions.
Ingestion points: Data is ingested via agent-browser snapshot and agent-browser get text as seen in capture-workflow.sh.
Boundary markers: No explicit boundary markers or instruction-ignoring warnings are present in the provided templates.
Capability inventory: The tool has broad capabilities including clicking, filling forms, executing JavaScript via eval, and writing files (screenshots, PDFs, and state files).
Sanitization: There is no evidence of sanitization or filtering of the extracted web content before it is processed by the agent.

agent-browser