autonomous-agent-patterns

Design patterns for building autonomous coding agents with tool integration, permissions, and human oversight. Design patterns for building autonomous coding agents, inspired by Cline and OpenAI Codex . Use this skill when: - Core agent loop architecture with think-decide-act-observe cycle; supports multi-model selection for different task types - Tool design patterns including file operations, code understanding, terminal execution, and browser automation with schema-based definitions - Permission system with four levels (auto, ask-once, ask-each, never) and risk assessment for approval workflows - Sandboxing and safety patterns including path validation, command whitelisting, and isolated execution environments - Context management with checkpoint/resume for long-running tasks and MCP (Model Context Protocol) server integration for dynamic tool discovery - Building autonomous AI agents - Designing tool/function calling APIs - Implementing permission and approval systems - Creating browser automation for agents - Designing human-in-the-loop workflows - Cline - OpenAI Codex - Model Context Protocol - Anthropic Tool Use - Use this skill only when the task clearly matches the scope described above. - Do not treat the output as a substitute for environment-specific validation, testing, or expert review. - Stop and ask for clarification if required inputs, permissions, safety...