Agent Skills › jjongguet/oh-my-product › ralplan

ralplan

GitHub

Ralplan 技能通过规划者、架构师和批评者的多角色迭代共识，为复杂任务和高危工作生成结构化执行计划。支持标准流程及包含预-mortem分析的深思模式，确保方案技术可行、风险可控且依赖无环。

skills/ralplan/SKILL.md jjongguet/oh-my-product

Trigger Scenarios

任务复杂度较高，单次规划不足需要架构决策前的结构化辩论高危工作需要事前分析和回滚策略

Install

npx skills add jjongguet/oh-my-product --skill ralplan -g -y

More Options

Use without installing

npx skills use jjongguet/oh-my-product@ralplan

指定 Agent (Claude Code)

npx skills add jjongguet/oh-my-product --skill ralplan -a claude-code -g -y

安装 repo 全部 skill

npx skills add jjongguet/oh-my-product --all -g -y

预览 repo 内 skill

npx skills add jjongguet/oh-my-product --list

SKILL.md

Frontmatter

{
    "name": "ralplan",
    "aliases": [
        "\/ralplan",
        "consensus plan",
        "consensus planning"
    ],
    "description": "Iterative planning with structured consensus — planner, architect, and critic converge on an approved plan.",
    "primaryRole": "orchestrator"
}

Ralplan Skill (oh-my-product)

Consensus-based planning that iterates until planner, architect, and critic agree.

Use when

The task is complex enough that a single planning pass is insufficient
Architecture decisions need structured debate before execution
High-risk work requires pre-mortem analysis

Workflow

Standard (default)

Planner drafts execution plan with task decomposition
Architect reviews for technical soundness and boundary issues
Critic challenges for completeness, risks, and missing edge cases
Iterate until all three agents converge (max 3 rounds)
Output: approved plan synced to /omp:taskboard

Deliberate (--deliberate flag, for high-risk work)

Adds to standard workflow:

Pre-mortem: "Assume this plan failed — what went wrong?"
Test strategy: Unit, integration, e2e, and observability planning
Rollback plan: How to undo if things go wrong
Extra iteration round allowed (max 4 total)

Convergence criteria

All three agents approve the plan
No unresolved HIGH severity concerns
Acceptance criteria are testable
Dependencies are acyclic

Output format

## Plan: [Title]
### Phases
Phase 1: [tasks, owners, dependencies]
Phase 2: ...

### Consensus
- Planner: APPROVE
- Architect: APPROVE (with note: ...)
- Critic: APPROVE

### Risk Register
| Risk | Impact | Mitigation |