zuoyebang/aiweave

🌟 For the latest content, please refer to the new homepage

AIWeave has moved into a brand-new main warehouse for ongoing development, with the latest version and subsequent evolutions now unified here

by XuRuibo · Continuous updates · Welcome to Star ⭐ and follow

---

> Author: XuRuibo <[email protected]>

Applicable to: Go backend microservices based on Gin / GORM / cobra / sarama Target audience: Project architects, AI agents (such as Claude Code)

AIWeave (AI Weaving)—The name echoes both sides of the norm:

• Weaving: Code ↔ and documentation are continuously synchronized bidirectionally, never breaking apart
• AI: All weaving actions are executed by AI according to standards, without relying on human self-awareness

*docs/ is the complete blueprint of the system, and .claude/skills/ is the AI’s operation manual. With just these two, AI can rebuild the entire project from scratch. *

AIWeave puts this assertion into reality: High-performance architecture methodology + directory structure specification + Documentation content specification + Skill step-by-step specification + Workflow + Self-checklist + Replicable Skeleton, allowing any compliant Go backend project to enjoy the triple benefits of “AI-native + AI rebuildable + high-performance architecture.”

Three Pillars — AIWeave not only “weaves documents after the fact” but also participates in architectural decision-making:

• design-spec/ High-Performance Architecture Methodology — How to Make Architectural Decisions in Response to Requirements (Technical Solutions / TRD Generation Reference)
• docs-spec/ Document Representation Norms — How to Write Decisions** (Precise to Reconstructable Code)
• skills-spec/ Skill Execution Specification — AI How to generate code based on documentation

The three are orthogonal: design handles “how to choose” · docs manages “how to write” · skills manages “how to do”.

---

1. What problems does this set of standards aim to solve?

The traditional relationship between code and documentation in traditional backend projects is: one person writes code → someone writes documentation (usually lagging or missing).

Under this paradigm:

• When refactoring occurs, documents cannot keep up with code drift;
• New members (whether human or AI) cannot quickly form a mental model of “why it is designed this way” when reading code;
• When rebuilding the same system on a new tech stack or platform, the design must be redone—the documentation does not retain executable level accuracy.

AI-native + AI rebuildable Paradigm reversal reverses this relationship:

Core assertion:

**docs/ is the complete blueprint of the system, and .claude/skills/ is the AI’s operating manual. With just these two, AI can rebuild the entire project from scratch. **

To achieve this assertion, the document must satisfy two hard constraints:

1. Code and documentation are extremely consistent — Any change in direction is forcibly synchronized to the other side;