[IMPORTANT] Use TaskCreate to break ALL work into small tasks BEFORE starting — including tasks for each file read. This prevents context loss from long files. For simple tasks, AI MUST ask user whether to skip.
Prerequisites: MUST READ .claude/skills/shared/understand-code-first-protocol.md before executing.
- docs/project-reference/domain-entities-reference.md — Domain entity catalog, relationships, cross-service sync (read when task involves business entities/models)
Skill Variant: Variant of /code — parallel phase execution from a plan.
Quick Summary
Goal: Execute implementation phases from an existing plan using parallel fullstack-developer subagents.
Workflow:
-
Load — Read the implementation plan and identify parallel phases
-
Dispatch — Launch subagents per phase with strict file ownership
-
Merge — Integrate results and verify
Key Rules:
-
Requires an existing plan file as input
-
Each subagent owns specific files; no cross-boundary edits
-
Sequential phases must wait for dependencies
Be skeptical. Apply critical thinking, sequential thinking. Every claim needs traced proof, confidence percentages (Idea should be more than 80%).
Execute plan: $ARGUMENTS
IMPORTANT: Activate needed skills. Ensure token efficiency. Sacrifice grammar for concision.
Workflow
- Plan Analysis
-
Read plan.md from given path
-
Check for: Dependency graph, Execution strategy, Parallelization Info, File Ownership matrix
-
Decision: IF parallel-executable → Step 2A, ELSE → Step 2B
-
External Memory: Re-read any .ai/workspace/analysis/ files referenced in the plan before dispatching to parallel agents.
2A. Parallel Execution
-
Parse execution strategy (which phases concurrent/sequential, file ownership)
-
Launch multiple fullstack-developer agents simultaneously for parallel phases
-
Pass: phase file path, environment info, file ownership boundaries
-
Wait for parallel group completion, verify no conflicts
-
Execute sequential phases (one agent per phase after dependencies)
-
Proceed to Step 3
2B. Sequential Execution
Follow ./.claude/workflows/primary-workflow.md :
-
Use main agent step by step
-
Read plan.md , implement phases one by one
-
Use project-manager for progress updates
-
Use ui-ux-designer for frontend
-
Run type checking after each phase
-
Proceed to Step 3
- Testing
-
Use tester for full suite (NO fake data/mocks)
-
If fail: debugger → fix → repeat
- Code Review
-
Use code-reviewer for all changes
-
If critical: fix → retest
- Project Management & Docs
- If approved: project-manager
- docs-manager in parallel (update plans, docs, roadmap)
- If rejected: fix → repeat
- Onboarding
- Guide user step by step (1 question at a time)
- Final Report
-
Summary, guide, next steps
-
Ask to commit (use git-manager if yes)
Examples:
-
Parallel: "Phases 1-3 parallel, then 4" → Launch 3 agents → Wait → Launch 1 agent
-
Sequential: "Phase 1 → 2 → 3" → Main agent implements each phase
IMPORTANT Task Planning Notes (MUST FOLLOW)
-
Always plan and break work into many small todo tasks
-
Always add a final review todo task to verify work quality and identify fixes/enhancements