orchestrating-tasks

Multi-step AI workflows are coordinated by routing tasks to an appropriate skill chain based on task characteristics like research, planning, implementation, or review. This routing ensures state is tracked and checkpoints are observed without writing code or making commits directly.

Yes, independent sub-tasks can be executed in parallel to improve throughput while maintaining dependency integrity. The orchestration logic manages this parallelization alongside the overall plan state to ensure workflow coherence.

Plan state is managed using brief and progress documents, enabling workflow resumption and checkpointed decision points. This document-based approach tracks progress across the skill chain without requiring direct code edits.

Yes, explicit user approvals are required, and the system prevents direct code edits or commits without user authorization. This ensures safe execution across planning, research, implementation, and review phases.

The best way to route an AI task is to provide an initial request and let the orchestrator analyze its characteristics to assign it to the correct skill chain. It automatically manages the progression through research, planning, implementation, and review.