Oh-My-DataStructer (OMD)

Data structures course assignment harness for Claude Code. From assignment description to submission package.

What is this?

OMD is a Claude Code plugin that provides a complete workflow for data structures course assignments. Give it your assignment description, and it will guide you through analysis, planning, implementation, testing, visualization, report generation, and submission packaging.

Features

• Auto-scaffolding: Detects DS topic, recommends language, creates project structure
• Architecture Blueprint: Decomposes problems into modules with basic/advanced/innovative methods (Natureflow-style)
• Multi-round Planning: gstack-style review rounds with interactive decision points
• 3D Visualization: Three.js interactive visualization with Glassmorphism HUD
• Chinese Academic Report: Word document with correct formatting (SimHei/SimSun/TNR, 1.5x spacing)
• Auto-Research: Karpathy-style iterative optimization with AHP/entropy-weight metric design
• Submission Packaging: Correct naming convention, zip with all deliverables

Installation

# Claude Code marketplace
claude plugin install mythrise/oh-my-datastructer

# Manual
git clone https://github.com/mythrise/oh-my-datastructer.git ~/.claude/plugins/oh-my-datastructer

(codex like this)

Quick Start

/ds:init "哈夫曼编码与文件压缩"
/ds:get-information
/ds:plan
/ds:implement
/ds:test
/ds:visualize
/ds:report
/ds:package

Supported Assignment Types

Type	Examples	Default Language
Compression/Encoding	Huffman, LZ77, arithmetic coding	Python
Sorting	Quicksort, mergesort, heapsort comparison	Python / C++
Trees	BST, AVL, Red-Black, B-tree	Python / C++
Graphs	Dijkstra, Prim, Kruskal, topological sort	Python
Hash Tables	Open addressing, chaining, perfect hashing	Python / C
Linear Structures	Stack, queue, linked list applications	C / Python

Commands

Command	Description
/ds:init "题目"	Initialize project from assignment description
/ds:get-information	Decompose into modules, survey academic methods
/ds:plan	Multi-round interactive planning
/ds:implement	Implement core algorithm code
/ds:test	Run tests + baseline/ablation/comparison experiments
/ds:visualize	Generate Three.js 3D visualization
/ds:report	Generate Word experiment report
/ds:auto-research	Iterative optimization loop
/ds:package	Package for submission

Codex 适配

OMD 原生支持 Claude Code 的 Codex 模式。在实现阶段，独立模块可并行委托 Codex 完成，大幅提升开发速度。

使用方式

方式一：在 /ds:implement 中自动委托

/ds:implement 会自动识别可并行的独立模块，将其委托给 Codex 执行。无需额外配置。

方式二：手动委托特定模块给 Codex

# 在 Claude Code 中使用 Codex 实现单个模块
/ds:implement --codex "core/encoder.py"

# 或直接在对话中指定
请使用 Codex 实现 src/core/encoder.py 中的编码器模块

方式三：Codex CLI 独立使用

# 将 OMD 作为 Codex 的上下文
codex --context .claude/plugins/oh-my-datastructer "实现 AVL 树的左旋和右旋操作"

Codex 最佳实践

• 适合委托: 核心算法实现、测试用例生成、基准测试脚本、数据导出器
• 不适合委托: 多轮规划讨论（/ds:plan）、创新点设计、报告撰写
• 并行策略: 当任务树中存在 ≥2 个无依赖模块时，自动触发 Codex 并行实现
• 质量门禁: Codex 产出自动经过单元测试验证，测试不通过则回退重试

与 Claude Code Agent 协作

┌─────────────────────────────────────────────┐
│  Claude Code (主控)                          │
│  ├─ /ds:plan        → 交互式规划             │
│  ├─ /ds:implement   → 分析依赖图             │
│  │   ├─ Codex #1    → 模块 A (并行)          │
│  │   ├─ Codex #2    → 模块 B (并行)          │
│  │   └─ Claude      → 模块 C (有依赖，串行)   │
│  ├─ /ds:test        → 集成验证               │
│  └─ /ds:report      → 报告生成               │
└─────────────────────────────────────────────┘

License

MIT