ai-engineering-system

📁 xcke/x-ai-system 📅 8 days ago
2
总安装量
1
周安装量
#64552
全站排名
安装命令
npx skills add https://github.com/xcke/x-ai-system --skill ai-engineering-system

Agent 安装分布

replit 1
amp 1
opencode 1
kimi-cli 1
github-copilot 1

Skill 文档

AI-Assisted Engineering System

Transform from passive assistant to active autonomous engineer through Context Engineering—storing product state, design, and architecture in structured Markdown files that can be read, reasoned about, and updated.

The .ai Context Engine

All intelligence lives in a hidden .ai directory—the Single Source of Truth (SSOT). A CLAUDE.md constitution in the project root ties everything together.

File Function Stage
CLAUDE.md The Constitution. Project-level AI instructions All
.ai/PRD.md The Why. Requirements, personas, metrics Planning
.ai/VIEWS.md The Vision. UI layouts, routes, mental models Design
.ai/SCHEMA.md The Structure. DB tables, API contracts, types Architecture
.ai/DESIGN_SYSTEM.md The Style. UI tokens, component patterns Implementation
.ai/TESTING.md The Gate. Verification protocols Verification
.ai/ROADMAP.md The State. Task list and sprint backlog Management
.ai/DECISIONS.md The Memory. Architecture Decision Records Documentation

Workflow Commands

Use these commands to operate the AI Engineering System:

Command Trigger Purpose
/x-feature New project/feature Plan features and roadmap like discussing with a product owner
/x-ask-architect Architecture decisions Discuss and brainstorm architecture with diagrams and trade-offs
/x-ask-ui-ux-designer UI/UX decisions Discuss layouts, views, design systems, and user flows
/x-plan Start of session Read context, identify next task, create execution plan
/x-build After plan approval Implement code following schema and design constraints
/x-verify After coding Run lint, build, test verification
/x-docs End of session Update ROADMAP.md and DECISIONS.md
/x-discover Existing project Reverse engineer codebase into .ai context files

Typical Session Flow

  1. Start: Run /x-plan to read context and get next task
  2. Approve: Review and approve the execution plan
  3. Build: Run /x-build to implement the task
  4. Verify: Run /x-verify to check lint, build, tests
  5. Document: Run /x-docs to update roadmap and decisions

For Existing Projects

Run /x-discover to bootstrap the .ai/ context system on established codebases through systematic reverse engineering. This will:

  • Scan project structure and identify framework/tools
  • Extract schema from database definitions
  • Detect UI patterns and design tokens
  • Generate draft context files with discovery markers
  • Guide you through review and confirmation

Initialization

When .ai/ directory doesn’t exist, create the context system:

mkdir -p .ai
  1. Generate CLAUDE.md in the project root (the constitution)
  2. Create each .ai/ file using the templates in references/templates.md

The CLAUDE.md file instructs Claude how to operate on this project and points to the .ai/ context files as the Single Source of Truth.

Resources

  • references/templates.md – Complete templates for all context files with examples (including CLAUDE.md)
  • references/brownfield-discovery.md – Detailed guide for reverse engineering existing codebases