AgentSkillsCN

ralph-deep-init

通过多阶段的架构分析，构建全面的项目待办事项清单。划分为六大功能模块，并为每个模块生成详细的任务列表。适用于需要为复杂项目打造庞大而条理清晰的待办事项清单时使用。

SKILL.md
--- frontmatter
name: ralph-deep-init
description: Build a comprehensive project backlog through multi-stage architectural analysis. Creates 6 functional groups and generates detailed tasks for each. Use when you need a large, well-organized backlog for complex projects.

Ralph Deep Initialization

This skill performs a comprehensive, multi-stage project initialization that generates a large, well-organized backlog by breaking down the project into functional architectural groups.

When to Use This Skill

Use ralph-deep-init instead of ralph-initialize when:

  • •Building a complex project with many features
  • •Need 20-40+ detailed tasks in the backlog
  • •Want tasks organized by architectural concerns
  • •Project has multiple distinct functional areas

Prerequisites

  • •Git repository initialized
  • •No existing prd.json (or user approves overwrite)
  • •Clear understanding of project requirements

Multi-Stage Process

Phase 1: Architecture

Goal: Initialize the project structure and define the architectural map.

  1. •

    Analyze Request

    • •Ask: "What do you want to build?" if not already provided
    • •Gather comprehensive project requirements
  2. •

    Create Foundation Files

    • •progress.md - Standard Ralph progress log
    • •.windsurf/rules/tech-stack.md - Tech stack & conventions
    • •init_progress.txt - Temporary initialization tracking
  3. •

    Generate Architecture

    • •Identify 6 distinct functional groups needed for this project
    • •Examples: Auth, Database, API, Frontend, Testing, DevOps
    • •Create groups.json containing ONLY a JSON array of these string names
    • •Format: ["Group1", "Group2", "Group3", "Group4", "Group5", "Group6"]

Phase 2: Expansion (6 Iterations)

Goal: Generate detailed tasks for each functional group.

For each of the 6 groups (iterate 1 through 6):

  1. •Select Target: Pick group N from groups.json
  2. •Generate Tasks: Create 3-5 detailed implementation tasks for this group
  3. •Write to File: Save to partial_N.json

Task Template:

Use .windsurf/skills/ralph-deep-init/task-examples.json as the reference schema for each partial_N.json.

Phase 3: Assembly

Goal: Compile the final Product Requirements Document.

  1. •

    Create PRD

    • •Create prd.json (use .windsurf/skills/ralph-initialize/prd-template.json as reference)
    • •Read all partial_*.json files
    • •Merge all tasks arrays into the backlog field
    • •Ensure valid JSON structure
  2. •

    Cleanup

    • •Delete groups.json
    • •Delete all partial_*.json files
    • •Keep init_progress.txt for reference
  3. •

    Report

    • •Announce: "Initialization Complete. Backlog created with [X] tasks."
    • •Summarize task breakdown by group

Supporting Resources

Reference these files:

  • •architecture-examples.md - Sample functional group breakdowns
  • •task-examples.json - Well-formed task examples
  • •groups-template.json - Template for groups array

Success Criteria

Deep initialization is complete when:

  • •✅ prd.json exists with 20+ tasks
  • •✅ Tasks are organized by functional groups
  • •✅ All tasks have passes: false
  • •✅ Each task has clear acceptance criteria
  • •✅ progress.md exists
  • •✅ Temporary files are cleaned up

Tips

  • •Choose groups that represent distinct technical concerns
  • •Ensure tasks within a group are cohesive
  • •Order tasks to respect dependencies
  • •Include infrastructure/testing as separate groups
  • •Aim for 3-5 tasks per group (total: 18-30 tasks)