March 21, 2026 • 6 min read

AI SDLC Scaffold: The Claude Code Template That Organizes Your Entire Project

If you use Claude Code (or any AI coding agent), you know the problem: your project starts clean, then chaos creeps in. Requirements get lost in Slack. Decisions get made in conversations but never documented. The AI forgets context after a few files.

Enter AI SDLC Scaffold — a free repository template that structures your entire software development lifecycle around AI-first workflows. Four phases. Built-in skills. Everything in-repo.

What It Actually Does

This isn't another tool or library. It's a starting point. You clone it, fill in your project details, and let Claude Code do the heavy lifting within a structured framework.

The scaffold organizes your project into four phases:

• 1-Objectives — What and Why. Stakeholders, goals, user stories, requirements, assumptions, constraints.
• 2-Design — How. Architecture, data models, API specs, decision records.
• 3-Code — Build. Task tracking, component directories, implementation guidelines.
• 4-Deploy — Ship. Runbooks, operational procedures.

Each phase has its own CLAUDE.[phase].md file with instructions, artifact indexes, and context. The AI always knows where it is and what's next.

Why It Matters for Indie Hackers

Most of us don't need enterprise SDLC processes. But we do need:

• Context that survives sessions — No more re-explaining project goals every time you start coding.
• Decisions that get recorded — Why did you choose PostgreSQL over Mongo? It's in decisions/.
• Token-efficient prompts — Hierarchical instructions mean the AI loads only what it needs.
• Supervision, not micromanagement — You steer at a high level; the AI executes within patterns.

The four core principles behind this scaffold:

• AI-first development model — AI does the work, you supervise and define objectives.
• Everything-in-repo — Requirements, architecture, decisions, and tasks live alongside source code.
• Context-window efficiency — Hierarchical instructions minimize token usage.
• Decision capture — Agents decide autonomously; all decisions are recorded for review.

Built-In Skills That Actually Help

The scaffold includes Claude Code skills (the /command style automation):

• /SDLC-init — Guided project initialization
• /SDLC-elicit — Requirements elicitation
• /SDLC-design — Design documents
• /SDLC-decompose — Component identification
• /SDLC-implementation-plan — Task generation
• /SDLC-execute-next-task — Task execution
• /SDLC-fix — Bug fixes and ad-hoc changes
• /SDLC-status — Project dashboard

You don't need to remember prompts. You just run the skill, and Claude Code knows exactly what to do in each phase.

How to Use It

# Quick start with degit
npx degit pangon/ai-sdlc-scaffold my-project
cd my-project
rm -f CONTRIBUTING.md CONTRIBUTORS.md LICENSE NOTICE RATIONALE.md README.md
git init && git commit -m "Initial scaffold"

# Then run the initialization skill in Claude Code
/SDLC-init

Or manually copy the files, remove the .git/ directory, and initialize your own repo.

What This Means for You

If you've ever:

• Lost track of what the AI was supposed to build
• Forgot why you made certain architectural choices
• Wasted tokens re-explaining project context
• Struggled to hand off a project to another developer (or AI)

— this scaffold solves that. It's not over-engineering. It's the minimum structure that makes AI-assisted development actually scalable.