The Agent Skills Directory

[COMMAND_EXECUTION]: The provided scripts (runner.js and test-suite.sh) execute the agent-browser CLI using system shell commands and include an eval command for browser-side JavaScript execution. This is the intended behavior for an automation tool and is considered safe for its primary purpose of testing.
[EXTERNAL_DOWNLOADS]: The skill documentation includes instructions to install the agent-browser package from npm and download Chromium via agent-browser setup. These resources are provided by Vercel Labs, a trusted organization, and follow standard software installation patterns.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it processes data from external websites. \n * Ingestion points: Content is retrieved via snapshot, text, and title commands in SKILL.md and runner.js.\n * Boundary markers: The scripts do not use specific delimiters to separate website content from agent instructions.\n * Capability inventory: The skill can execute shell commands via execSync and JavaScript via agent-browser eval.\n * Sanitization: There is no sanitization of the retrieved web content before it is used in test assertions.

e2e-agent-browser