$ralplan
Planning skill that loops until planner, architect, and critic all agree
TL;DR
$ralplan has planner, architect, and critic — three agents — iteratively review a plan until they all agree. It's the step that locks in tradeoffs right before implementation.
How to invoke
$ralplan "approve the safest implementation path"Also activated by magic keywords ralplan, consensus plan.
When to use it
| Good fit | Avoid |
|---|---|
| Work involving architecture decisions | Work that only needs a quick draft |
| When the wrong direction is expensive | Small features, short-term tasks |
| Need to explicitly review multiple tradeoffs |
Internal flow
- planner — drafts implementation steps and order.
- architect — reviews boundaries and interfaces from a system design perspective.
- critic — flags risks and missing pieces.
- If there are objections, revise and review again. When all agree, finalize the plan.
Where it fits in the standard flow
$deep-interview "..." ← confirm requirements
$ralplan "..." ← consensus here
$ralph "..." ← execute--deliberate option
For high-risk work, add --deliberate. It adds a pre-mortem, unit/integration/e2e test plans, and operational impact analysis.
$ralplan --deliberate "plan replacement of legacy auth system"Difference from $plan
$plan lets the planner draft something solo, fast. $ralplan trades that speed for agreement. It runs three agents and drives them to a consensus, which is what you want when a stable plan matters more than a quick one.
Related skills
$deep-interview— clarify requirements before planning$ralph— execute the consensus plan$plan— solo planning (quick draft)