Orchestrates parallel execution of dependent tasks across multiple subagents using markdown plan files.

• Parses task dependencies from plan markdown and launches unblocked tasks in parallel waves, ensuring correct execution order
• Provides subagents with structured task context including descriptions, acceptance criteria, validation steps, and dependency information
• Enforces test-driven development (RED phase first) with fallback to documented non-testable verification for tasks that don't support automated testing
• Validates task completion through test evidence or explicit verification, commits work locally per task, and updates plan documentation with logs before advancing to the next wave
• Supports selective task execution via task ID filtering while automatically including required dependencies