autonomous-agent-patterns

SUSPICIOUS. The skill is coherent as an agent-design guide, not overtly malicious, and it lacks suspicious installers or clear exfiltration endpoints. However, it teaches high-impact patterns: autonomous command/browser actions, ingestion of untrusted external content into LLM context, and dynamic creation/loading of MCP tools, which materially increase implementation risk for an AI agent.