Released Common Rules for AI-Assisted Development
Best Practices Born from Real Projects
Published common rules extracted from multiple AI collaboration projects on Gist. Systematized practical know-how including proactive questioning, work keywords, and communication principles.
Why We Published Common Rules
After about two weeks of AI-assisted development using Claude Code across multiple projects, we've identified common patterns and effective methods.
Today, we're publishing this knowledge as "Common Rules for AI-Assisted Development."
🔗 AI Collaboration Common Rules v2.0
What's Included
1. 🎯 Basic Principles
Four crucial principles for AI collaboration:
- Information Priority - System reality > Documentation
- Thorough Self-Investigation - Research before asking
- Balanced Proactivity - Don't proceed alone, but do suggest
- Proactive Questioning - Confirm rather than assume
2. 🤔 Proactive Questioning Principles
We particularly focused on "proactive questioning":
【5 Situations Where Questions Are Essential】
1. Large impact scope (5+ files)
2. Data structure changes
3. Implementation differing from existing patterns
4. Performance/security implications
5. Similar past failuresProper timing of AI questions prevents major mistakes and leads to better solutions.
3. 📝 Work Management
The "keyword" system born from actual development:
- "Mid-close" = Document update + Daily report update
- "Close" = Mid-close + git add + commit + push
These concise commands dramatically improved routine work efficiency.
4. 💬 Communication
- The 4-line rule:
- AI responses within 4 lines (excluding code/tools)
- Avoid verbose explanations, deliver essential information only
✅ Good: "Completed. Updated 3 files."
❌ Bad: "Yes, I understand. Let me start working on this... (long explanation)"Born from Real Projects
These aren't theoretical rules. They emerged from actual projects:
- GIZIN Corporate Site - Large-scale Next.js + TypeScript project
- StressCheckHarmony - SaaS application under development
And more, through continuous trial and error.
Why Publish on Gist?
1. Easy Version Control
Leverage GitHub features to track rule evolution.2. Multi-Project Reference
Using Claude Code's @import feature enables centralized management of common rules.
# Project Name Operation & Improvement Guide
@../COMMON_RULES.md
## 📌 Project-Specific Rules
(Project-specific content follows)- Benefits:
- Updates to common rules instantly reflected across all projects
- Reliably loaded when Claude Code starts
- Flexible placement according to project structure
3. Community Contribution
AI-assisted development is still new. As pioneers, we want to share knowledge for industry-wide advancement.Work in Progress
This is a "living document":
- v2.0 - Added proactive questioning principles
- v1.0 - Initial release
We'll continue improving through practice and welcome feedback.
For Your Projects
Use these rules as-is or customize for your projects.
The key is "documenting rules" for AI collaboration, enabling:
- Cross-session consistency
- Team-wide quality standards
- Rapid onboarding of new members
Conclusion
AI-assisted development is still evolving. That's why sharing knowledge among practitioners is crucial for collective growth.
We hope these common rules help your AI collaboration journey.
Feedback welcome via Gist comments or Developer's X!
- ---