A methodology for building production-ready apps with AI — systematically, not randomly.
A structured approach to AI-assisted development that prevents the common failure pattern: spending thousands on tokens and ending up with broken code.
The process:
Copy project-templates/ into your repo. It includes:
.clinerules and CLAUDE.md with strict quality rulesproject-templates/ # Drop-in files for any project
├── README.md # Project README template
├── ARCHITECTURE.md # System design template
├── LEARNINGS.md # Gotchas log template
├── .clinerules # Cline quality rules template
├── CLAUDE.md # Claude Code quality rules template
├── TASK_TEMPLATE.md # Task specification format
├── SPRINT_PLAN_TEMPLATE.md # Sprint planning format
├── SPRINT_RULES.md # Sprint sizing and task rules
└── prompts/ # Copy-paste prompts for each phase
├── 01-initial-brainstorm.md
├── 02-generate-foundation-docs.md
├── 03-generate-task-docs.md
├── 04-new-feature-brainstorm.md
├── 05-fix-and-debug.md
├── 06-context-rescue.md
└── 07-phase-audit.md
docs/ # The methodology guide (VitePress)
├── introduction.md
├── part-1/ # Foundation
├── part-2/ # Pre-Development
├── part-3/ # Execution
├── part-4/ # Quality
├── part-5/ # Advanced
├── part-6/ # Resources & templates
└── .vitepress/ # Site config
The VH Conference Toolkit was built using this methodology. Browse its repo to see thorough architecture docs, strict development rules, sprint-based task specs, and ADRs in action.
This methodology built RISE — an Electron desktop app — in 4 weeks for ~$400 in tokens. Production-ready, documented, maintainable.
Major update complete. Core methodology revised to reflect current best practices. Project templates and prompt library added.