The Agent Skills Directory

[PERSISTENCE]: The skill provides a launchd configuration file (is.ars.chrome-cdp.plist) and setup instructions to install it into ~/Library/LaunchAgents/. This ensures the Chrome browser instance automatically starts upon user login.
[EXTERNAL_DOWNLOADS]: The scripts/install-cft.sh script uses npx @puppeteer/browsers to download the 'Chrome for Testing' binary from Google's official infrastructure. This is a standard and documented behavior for the skill's stated purpose.
[INDIRECT_PROMPT_INJECTION]: The skill enables browser automation using Playwright, creating a surface where the agent ingests untrusted data from external websites. Malicious content on those sites could attempt to influence the agent's behavior through snapshots or text extraction.
Ingestion points: Tool outputs from browser_snapshot, browser_tabs, and text content extracted via innerText (documented in SKILL.md).
Boundary markers: None explicitly defined in the skill's logic to distinguish between instructions and data.
Capability inventory: The skill includes scripts that execute shell commands (launchctl, xattr), file system writes (~/.local/bin/ symlinks), and network operations via a controlled browser.
Sanitization: No specific sanitization or filtering is implemented for content retrieved from the browser.
[COMMAND_EXECUTION]: The provided utility scripts execute various shell commands to manage the environment, including launchctl for service management, xattr to clear macOS quarantine flags, and npx for package execution.

chrome-browser