English | 中文说明

Architect Fullstack Skill

A practical AI skill pack for architects, tech leads, and fullstack engineers.

This repository helps you make AI behave more like a real engineering partner instead of a generic chatbot.

It focuses on:

• architecture thinking
• Go backend engineering
• API design
• database design and optimization
• review and refactoring
• release and rollback awareness
• Web frontend
• mini program
• app client collaboration

---

Why this exists

Most prompts are too generic.

They can answer questions, but they usually fail to behave like a real:

• architect
• core engineer
• reviewer
• release owner

This skill pack is designed to close that gap.

It gives you a reusable structure for making AI work with:

• clearer engineering judgment
• better architecture trade-off analysis
• stronger implementation guidance
• better review quality
• stronger release-risk awareness

---

What makes it different

Instead of providing only one large prompt, this repository is organized as a layered system:

• core/ for shared engineering mindset
• adapters/ for Claude / Cursor specific behavior
• rules/ for specialized engineering rules
• workflow/ for execution process and review flow
• examples/ for reusable task prompts
• packs_rendered/ for ready-to-use final prompts

That means you can either:

• use it directly
• or maintain and evolve it like a real prompt engineering system

---

Repository structure

public_skill/
├── core/
│   └── architect_fullstack_core.md
├── adapters/
│   ├── claude_adapter.md
│   └── cursor_adapter.md
├── rules/
│   ├── go_backend_engineering_rules.md
│   ├── frontend_engineering_rules.md
│   ├── web_frontend_rules.md
│   ├── miniprogram_rules.md
│   ├── app_client_rules.md
│   ├── database_engineering_rules.md
│   ├── api_design_rules.md
│   └── deployment_release_rules.md
├── workflow/
│   ├── architecture_workflow.md
│   ├── review_checklist.md
│   └── task_prompt_templates.md
├── examples/
│   ├── web_task_examples.md
│   ├── miniprogram_task_examples.md
│   └── app_client_task_examples.md
├── packs_rendered/
│   ├── architect_fullstack_claude_skill.md
│   ├── architect_fullstack_cursor_skill.md
│   ├── go_backend_engineering_skill.md
│   ├── api_design_skill.md
│   ├── web_frontend_skill.md
│   ├── miniprogram_skill.md
│   ├── app_client_collaboration_skill.md
│   ├── engineering_review_skill.md
│   └── engineering_release_skill.md

---

Quick start

Option 1: Install as Claude/Cursor Skill (Recommended)

# Clone or copy this repository
git clone <repo-url> architect-fullstack-skill
cd architect-fullstack-skill

# Run installation script
chmod +x install.sh
./install.sh

Choose your installation location:

• Claude: ~/.claude/skills/architect-fullstack-skill
• Cursor: ~/.cursor/skills/architect-fullstack-skill

After installation:

1. Restart Claude/Cursor
2. Go to Settings → Skills to load the skill
3. Use trigger words to activate:

Trigger Word	Description
@architect-fullstack	Full architect skill
@go-backend	Go backend focused
@api-design	API design focused
@web-frontend	Web frontend focused
@miniprogram	Mini program focused
@app-client	App client collaboration
@review	Code review focused
@release	Release assessment focused

Option 2: Use ready-made skills directly

If you want the fastest path, start with files in packs_rendered/.

Recommended entry points:

• Claude: packs_rendered/architect_fullstack_claude_skill.md
• Cursor: packs_rendered/architect_fullstack_cursor_skill.md
• Backend engineering: packs_rendered/go_backend_engineering_skill.md
• API design: packs_rendered/api_design_skill.md
• Web frontend: packs_rendered/web_frontend_skill.md
• Mini program: packs_rendered/miniprogram_skill.md
• App collaboration: packs_rendered/app_client_collaboration_skill.md
• Review: packs_rendered/engineering_review_skill.md
• Release: packs_rendered/engineering_release_skill.md

Option 2: Maintain the system yourself

If you want long-term maintainability, start from:

• core/architect_fullstack_core.md
• adapters/claude_adapter.md
• adapters/cursor_adapter.md
• rules/
• workflow/

---

Recommended usage patterns

For Claude

Use:

• packs_rendered/architect_fullstack_claude_skill.md

Best for:

• architecture design
• deep debugging
• trade-off analysis
• refactoring strategy
• high-risk review

For Cursor

Use:

• packs_rendered/architect_fullstack_cursor_skill.md

Best for:

• implementation
• local refactoring
• editor-centric iteration
• current-module optimization

For specialized tasks

Use the dedicated packs:

• Go backend: go_backend_engineering_skill.md
• API design: api_design_skill.md
• Web: web_frontend_skill.md
• Mini program: miniprogram_skill.md
• App client: app_client_collaboration_skill.md
• Review: engineering_review_skill.md
• Release: engineering_release_skill.md

---

Who this is for

This repository is a strong fit if you are:

• an architect
• a tech lead
• a senior Go backend engineer
• a fullstack developer
• an AI power user who wants better engineering outputs

---

Who this is not for

This repository is probably not a good fit if you want:

• a tiny one-line prompt
• beginner-level coding explanations only
• a framework-specific assistant only
• a project-specific internal rule set

This is a general public engineering skill pack, not a private company rule base.

---

Public scope

This public release intentionally excludes:

• private project structure
• internal business rules
• company-specific engineering process
• project-specific pitfalls and private context

That keeps this repository reusable and safe for public release.

---

Suggested release strategy

You can publish this repository in several ways:

Open source repository

Best for:

• GitHub
• public collaboration
• issue / PR based evolution

Prompt product / knowledge pack

Best for:

• Gumroad
• Notion
• blog download pack
• paid prompt bundle

Internal team starter kit

Best for:

• adapting the public version into a company-internal prompt system

---

Maintenance model

Recommended priority:

1. update core/
2. update adapters/
3. update rules/ and workflow/
4. regenerate or revise packs_rendered/

In other words:

• maintain the source layers first
• treat packs_rendered/ as the final consumable layer

---

Version

Current version: v1.0.1

See also:

• SKILL_MANIFEST.md
• CHANGELOG.md
• LICENSE

---

License

MIT