Project Scaffold

🏗️ Project Scaffold — AI Agent Config Generator

Generate all AI-agent config files for any full-stack project — in one go.

A cross-platform AI agent skill that interviews you about your project, then generates a complete set of CLAUDE.md, rules, agents, skills, commands, progress tracking, and reference docs. Every file is self-updating, every agent asks before assuming, and the system tracks what's done, what's next, and when to stop.

---

What It Does

Describe your project → the skill interviews you → generates everything:

.claude/
├── rules/           # code-style, testing, security, git-conventions
├── agents/          # code-reviewer, test-writer, debugger, architect
├── skills/          # new-feature, new-endpoint, new-component
└── commands/        # /new-feature, /fix-bug, /deploy-check, /status, /review
docs/agent/
├── architecture.md, database-schema.md, api-patterns.md
├── workflows.md, testing-strategy.md, deployment.md
└── progress.md      # living progress tracker
CLAUDE.md            # core memory — lean, <300 lines, self-updating

---

The 9 Phases

Phase	What Happens
1. Discovery Interview	Asks structured questions — never assumes stack or conventions
2. File Generation	Creates all config files in one batch
3. Self-Updating Docs	Every file has <!-- UPDATE_WHEN --> markers and changelogs
4. Sub-Agents	Code-reviewer, test-writer, debugger, architect — with tool access
5. Ask-Don't-Assume	Clarification protocol baked into every agent and CLAUDE.md
6. Next Step Engine	Suggests 2-3 prioritized next actions after every task
7. Progress Tracking	Milestones, task log, decisions, tech debt, open questions
8. Best Practices & Coaching	Contextual coaching + knows when to stop
9. /status & /review	Instant project health check and review checkpoints

---

Key Features

🎯 Interview First, Generate Second

The skill never assumes your stack. It asks tiered questions (Must Know → Should Know → Nice to Know) and blocks file generation until the essentials are gathered. If you've already described your project in conversation, it extracts what it can and confirms.

📊 Progress Tracking

A living docs/agent/progress.md tracks milestones customized to your project type:

Project Type	Example Milestones
SaaS / Dashboard	Auth, core API, user management, billing, testing, deploy
E-commerce	Product catalog, cart, checkout, payment, order management
API-only	Resource endpoints, validation, rate limiting, API docs
Content Platform	CMS, media upload, publishing workflow, search
Portfolio Site	Layout, content sections, contact form, SEO, analytics
Mobile Backend	Auth with device management, push notifications, real-time

Updated after every task. Read at the start of every session.

💡 Best Practices Coaching

20+ contextual suggestions across Security, Performance, Quality, Reliability, and Developer Experience — each with a specific trigger condition. Maximum one per task. Always skippable.

✅ Completion Awareness

Three levels of "done":

• Task-level → Stop, confirm, offer enhancements as options (not defaults)
• Feature-level → Suggest review checkpoint before next feature
• Project-level → Launch readiness summary when all milestones are complete

Anti-overbuilding rules prevent gold-plating, premature optimization, and feature creep.

🔄 Self-Updating Documents

Every generated file includes:

• <!-- UPDATE_WHEN: condition --> markers at key decision points
• Changelog table at the bottom
• Instructions in CLAUDE.md for the agent to scan for affected docs after every task

---

Usage

Trigger Phrases

Just describe your project naturally:

"I'm building a SaaS dashboard with Next.js, Prisma, and PostgreSQL. Generate the docs."

Or be explicit:

"Use the project-scaffold skill to set up agent config."

Other triggers: "scaffold the project", "create the MD files", "set up Claude Code config", "generate the docs"

After Setup

Command	What It Does
/status	Project health: milestones, last task, next step, tech debt
/review	Feature/milestone review checkpoint with code-reviewer agent
/new-feature	Full implementation workflow with planning and testing
/fix-bug	Structured debugging protocol
/deploy-check	Pre-deployment verification checklist

---

Installation

# Register the marketplace
/plugin marketplace add Imit96/imit-dev-plugins

# Install the skill
/plugin install project-scaffold@imit-dev-plugins

---

License

MIT

---

Built by Imit using Claude AI's skill system.