Plan Research

Create a planner-led orchestration strategy before execution.

The plan-research workflow creates the orchestration brain for complex runs.

How it works

  1. Reads your topic and constraints.
  2. Builds an agent sequence and handoff plan.
  3. Sets stop criteria (confidence threshold, contradiction targets).
  4. Defines verification checkpoints.

Real examples

/plan-research "safety and performance trade-offs in long-context RAG"
bohr plan-research "safety and performance trade-offs in long-context RAG"

Expected output

  • outputs/.plans/<slug>-orchestration.md

Includes: questions, agent order, loop policy, and run-next recommendation.

Customization

  • Set confidence target (for example, 0.8+)
  • Set contradiction tolerance (for example, zero critical contradictions)
  • Choose max loop count and verification strictness