Production AI Workflow
Claude Code SaaS Repository System
A structured repository workflow designed for scalable AI-assisted SaaS engineering.
24 rules12 prompts6 workflowsUpdated 3 days ago
Repository Context
Stack:
- Next.js App Router
- Supabase
- Stripe
- shadcn/ui
- pnpm
- Turborepo
Architecture:
- server-first
- feature-based
- shared UI primitives
Constraints:
- avoid duplicated hooks
- avoid mixed concerns
- keep PRs under 300 LOC
Repository Structure
/app /(dashboard) /(marketing) /api /components /ui /shared /features /auth /billing /dashboard /settings /analytics /notifications /lib /db /stripe /auth /validators /utils /prompts /rules /tests legacy/ deprecated/ README.md TODO.md .env.example pnpm-workspace.yaml turbo.json architecture.md migration-notes.md testing-guidelines.md memory.md rules.md
Workflow Steps
Step 1 — Repository Planning
Constraints:
- define feature boundaries
- avoid shared business logic
- keep folder naming predictable
Expected Output:
- repo structure
- feature folders
- architecture notes
Common Failures:
- oversized shared utils
- mixed feature concerns
- inconsistent naming
Step 2 — Feature Planning
Constraints:
- avoid duplicated hooks
- isolate business logic
- reuse validation schemas
Expected Output:
- feature folder
- server actions
- validation layer
Common Failures:
- repeated schemas
- client-side fetching
- oversized components
Step 3 — AI Implementation
Constraints:
- follow repo conventions
- reuse shared primitives
- preserve server-first architecture
Expected Output:
- predictable code structure
- isolated business logic
- reusable UI
Common Failures:
- inconsistent prompt output
- duplicated logic
- unstable patterns
Step 4 — PR Validation
Constraints:
- avoid mixed concerns
- keep PRs focused
- preserve architecture consistency
Expected Output:
- stable PR review
- predictable diffs
- maintainable repo history
Common Failures:
- oversized PRs
- inconsistent architecture
- duplicated patterns
Prompt
Refactor this feature following existing repo conventions.Requirements:- preserve server-first architecture- avoid duplicated validation logic- reuse shared UI primitives- keep business logic isolatedDo not:- introduce new folder structures- fetch inside client components- create oversized utility files
Team Workflow
1. feature planning
2. AI implementation
3. human review
4. PR validation
5. incremental refactor
All generated code must pass review before merge.
Common AI Engineering Failures
- duplicated hooks
- inconsistent naming
- mixed client/server logic
- oversized components
- scattered utilities
- repeated validation schemas
Migration Notes
*dashboard still uses old hooks
*billing refactor in progress
*auth migration planned
*validation migration partially completed
Recent Changes
// changelog
v0.4.2
- improved billing workflow
- updated repo constraints
- added testing standards
- reduced duplicated logic
v0.4.1
- updated auth workflow
v0.4.0
- introduced repo constraints