The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/with_server.py utility executes shell commands provided as arguments (such as npm run dev) to manage development server lifecycles during testing. It uses subprocess.Popen with shell=True, which is necessary for executing shell commands across different platforms. This is a standard pattern for development tooling.\n- [INDIRECT_PROMPT_INJECTION]: The skill interacts with external web applications, which constitutes an attack surface for indirect prompt injection.\n
Ingestion points: The agent reads rendered HTML and captures screenshots via Playwright MCP tools as documented in references/playwright-mcp-cheatsheet.md and SKILL.md.\n
Boundary markers: The instructions recommend waiting for specific application states or window flags (window.__TEST__.ready), which helps ensure stability but does not provide formal prompt isolation from untrusted page content.\n
Capability inventory: The skill includes scripts for shell command execution (scripts/with_server.py) and file writing (scripts/imgdiff.py).\n
Sanitization: No explicit sanitization of UI content or console logs is performed before the agent processes them.

frontend-testing