AGENTS.md Interactive Generator
Choose a stack and project boundaries to draft an editable
AGENTS.md. Before copying it to the project root, check the commands, paths, and permission requirements. Written rules work alongside sandbox controls, code review, and tests.
Real-time AGENTS.md Preview:
# 🤖 Codex Collaboration Protocol (CAP)
## 📌 Project Signature
- **Project Name**: my-saas-mvp
- **Target Architecture**: Next.js (App Router) + TypeScript
- **Before use**: Adapt commands, paths, and permissions to this repository. Written rules do not enforce sandbox access or replace review.
## Environment and Permission Guidance
- Standard development: verify install, test, and build commands; use actual permission settings for production and credential access.
## Engineering Rules
1. **🛑 Stop Infinite Loops (Prevent Loop)**: If compilation or test fails twice in a row, halt immediately, log the full error context, and switch to single-module diagnostic mode. Continuous guessing loops are strictly forbidden.
2. **🚫 Ban Fake Stubs (No Placeholders)**: No stubs like `// TODO: implement`, `pass`, or fabricated mock data. All code must be runnable.
3. **📦 Dependency Guard**: Never introduce heavyweight dependencies or third-party wrappers without explicit review. Prefer standard library solutions.
4. **Stack-Specific**: Components default to React Server Components; only add "use client" when interaction is required
5. **Stack-Specific**: Use Tailwind CSS; inline arbitrary style objects are prohibited
6. **Stack-Specific**: Strictly separate server envs from client NEXT_PUBLIC_ variables
## 🧪 Deliver with Tests (Verification Specs)
Before marking any task as complete, verify that the following commands exist in the project, run the applicable checks, and record their exit status:
```bash
# 1. Lint & Unit Tests
npm run lint && npm test
# 2. Production Build Verification
npm run build
```

