Code-MCP

# Adaptive Agent Workflow This workflow is designed for autonomous agents to systematically plan, execute, and verify complex tasks. ## 1. Planning Phase - [ ] **Analyze Request**: Understand the user's goal, context, and constraints. - [ ] **Explore Codebase**: Use `list_files`, `view_file_outline`, and `grep_search` to understand existing systems. - [ ] **Define Task**: Use `task_boundary` to set the `TaskName` and `TaskSummary`. - [ ] **Create Plan**: Use `plan_task` to break down the work into steps. - [ ] **Create Implementation Plan**: Create `implementation_plan.md` using the artifact format. - [ ] **Review**: Request user review of the plan using `notify_user` if the task is complex or risky. ## 2. Execution Phase - [ ] **Task Management**: Update `task.md` and `task_boundary` as you progress. - [ ] **Step-by-Step**: Execute one step at a time. - [ ] **Sequential Thinking**: Use `sequential_thinking` for complex logic or when stuck. - [ ] **Code Changes**: Use `replace_file_content` or `multi_replace_file_content` for edits. - [ ] **Validation**: Verify each change immediately after making it (e.g., compile, lint, run test). - [ ] **Artifact Updates**: Update `task_boundary` status frequently. ## 3. Verification Phase - [ ] **Comprehensive Testing**: Run relevant tests using `run_command` (e.g., `npm test`). - [ ] **Visual Verification**: Use `browser_subagent` or `generate_image` if applicable. - [ ] **Documentation**: Update `walkthrough.md` with proof of work (diffs, logs, screenshots). - [ ] **Final Review**: Self-correction using `reflect_on_code` or `act_as_reviewer`. ## 4. Completion - [ ] **Cleanup**: Remove temporary files or debug logs. - [ ] **Final Notification**: Use `notify_user` to inform the user of completion and point to the `walkthrough.md`. ## Decision Heuristics - **Complexity**: If > 3 files affected, create a detailed plan. - **Risk**: If deleting code or changing DB schema, require user approval. - **Uncertainty**: If unsure, prototyped first or ask the user.