maker-orchestrator

Multi-step task automation uses orchestration to decompose complex requests into parallel subtasks, execute them sequentially or in parallel, validate results, and recover from failures. This Skill detects engineering tasks automatically and runs the full MAKER pipeline—decomposition, planning, execution, validation, and review—with checkpoint recovery and caching for resilience.

Checkpointing preserves task progress and intermediate results between workflow steps, enabling fast recovery without re-running completed work. This Skill manages inter-step state passing, caches results, and coordinates Task tool execution across decomposed subtasks, so you resume from the last successful checkpoint on retry.

Parallel execution planning breaks complex tasks into independent subtasks that run concurrently, then merges results. This Skill orchestrates parallel planning stages, validates outputs across branches, applies voting or consensus where needed, and coordinates convergence before proceeding to the next workflow phase.

Red-flag validation checks each step's output against quality criteria and automatically retries up to three times on failure, with state preserved between attempts. This Skill applies validation voting across subagents, triggers retry loops on red flags, and manages recovery without manual intervention.

Task orchestration fits complex requests that require decomposition into subtasks, parallel planning, cross-step validation, and recovery from intermediate failures. Use it when tasks span multiple stages, need caching between runs, or benefit from voting and checkpointing—sequential execution suffices for linear, single-stage work.

This Skill coordinates Task tool execution natively and integrates with Claude's cache to reduce redundant computation. It manages state passing between tools, orchestrates parallel tool calls across subagents, and applies checkpoint recovery, so existing integrations work within the MAKER pipeline without modification.