harness-work

To automate Plans.md execution, you load the task plan and instruct the system to run pending tasks end-to-end. It interprets the plan, applies DoD validation, and generates sprint contracts to ensure reproducible results.

Multi-agent orchestration coordinates Lead, Worker, Advisor, and Reviewer roles to execute tasks in parallel. It ensures reproducible research workflows by automating review loops, change management, and generating rich completion reports.

Yes, Claude Code workflows support parallel task execution modes like Breezing. The system applies Definition of Done validation, coordinates multiple agents, and generates sprint contracts to guarantee verifiable task completion.

Yes, you can execute a single pending task or all remaining tasks end-to-end. The system automatically selects the appropriate mode, coordinating Solo, Parallel, Codex, or Breezing execution to handle the workload.

The best way to maintain traceability is through automated failure re-ticketing and rich completion reports. When DoD validation fails, the system automatically re-tickets the task to maintain audit traceability across the workflow.