The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions and automated commands for downloading official tools and libraries, such as the @anthropic/gws CLI tool, the ai-agent-skills installer, and various Node.js dependencies for Playwright and MCP integrations from the official npm registry.
[COMMAND_EXECUTION]: Multiple Python automation tools (e.g., pr_analyzer.py, gws_doctor.py, workspace_audit.py) use subprocess.run() to interface with standard CLI utilities like git and gws. Additionally, the playwright-pro sub-skill implements automated hooks that execute shell scripts for test validation and environment detection upon file operations.
[PROMPT_INJECTION]: The ai-security sub-skill documentation and its ai_threat_scanner.py tool contain lists of prompt injection signatures, such as 'ignore previous instructions' and 'DAN mode'. These are part of a defensive signature database used to detect and score injection risks in AI systems, rather than instructions intended to subvert the agent's behavior.

engineering-skills