playbooks

home / skills / yeachan-heo / oh-my-claudecode / ralplan

ralplan skill

/skills/ralplan

This skill orchestrates consensus planning by coordinating Planner, Architect, and Critic to deliver validated plans with user-approved execution.

npx playbooks add skill yeachan-heo/oh-my-claudecode --skill ralplan

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
1.1 KB
---
name: ralplan
description: Alias for /plan --consensus
---

# Ralplan (Consensus Planning Alias)

Ralplan is a shorthand alias for `/oh-my-claudecode:plan --consensus`. It triggers iterative planning with Planner, Architect, and Critic agents until consensus is reached.

## Usage

```
/oh-my-claudecode:ralplan "task description"
```

## Behavior

This skill invokes the Plan skill in consensus mode:

```
/oh-my-claudecode:plan --consensus <arguments>
```

The consensus workflow:
1. **Planner** creates initial plan
2. **User feedback**: **MUST** use `AskUserQuestion` to present the draft plan before review (Proceed to review / Request changes / Skip review)
3. **Architect** reviews for architectural soundness
4. **Critic** evaluates against quality criteria
5. If Critic rejects: iterate with feedback (max 5 iterations)
6. On Critic approval: **MUST** use `AskUserQuestion` to present the plan with approval options
7. User chooses: Approve, Request changes, or Reject
8. On approval: **MUST** invoke `Skill("oh-my-claudecode:ralph")` for execution -- never implement directly

Follow the Plan skill's full documentation for consensus mode details.

Overview

This skill is an alias that runs the Plan workflow in consensus mode: /oh-my-claudecode:plan --consensus. It orchestrates Planner, Architect, and Critic agents to produce a vetted plan and hands off execution to the ralph execution skill after user approval. Use it when you need iterative, team-style planning with explicit user reviews and automated quality gates.

How this skill works

Ralplan invokes the Plan skill with the --consensus flag, starting a loop where a Planner drafts a plan, the Architect checks architecture, and the Critic enforces quality criteria. The flow requires presenting the draft and final plan to the user via AskUserQuestion steps, supports up to five Critic-driven iterations, and—on approval—calls Skill("oh-my-claudecode:ralph") to perform execution. The skill never implements execution itself and follows the Plan skill’s consensus semantics.

When to use it

  • You need a multi-agent, iterative plan with explicit user review checkpoints.
  • Architectural validation and automated critique are required before execution.
  • You want a repeatable consensus process with a fixed iteration cap.
  • You must ensure human approval before any automated execution step.
  • Coordinating team-style reasoning across Planner, Architect, and Critic agents.

Best practices

  • Provide a clear, scoped task description to speed consensus and reduce iterations.
  • Expect and enforce AskUserQuestion prompts: present the draft, then present the approved plan.
  • Use concise review options (Proceed to review / Request changes / Skip review) to guide users.
  • Allow up to five iterations for Critic feedback; adjust task scope if iterations reach the cap.
  • Never implement execution in this skill; rely on the mandatory call to ralph for running changes.

Example use cases

  • Designing a new microservice: generate an initial plan, validate architecture, and collect approval before deployment.
  • Coordinating a multi-step refactor where quality checks and user sign-off are required.
  • Creating a release plan that needs iterative critique for test coverage, rollback strategy, and performance goals.
  • Preparing a complex infrastructure change where an architect must verify topology and a critic enforces standards.

FAQ

What happens if the Critic keeps rejecting the plan?

The workflow will iterate up to five times with feedback loops. If consensus is not reached after five iterations, revise the task scope or respond to the Critic’s guidance to break the cycle.

Can this skill execute changes directly?

No. On user approval the skill must invoke Skill("oh-my-claudecode:ralph") to perform execution; it does not implement execution itself.