Dan Trapp

Open-source infrastructure / Developer education

Claude Code Workflow

An open-source workflow system for running Claude Code with specs, repo rules, PR discipline, and human review built in.

A public workflow kit for solo builders, teams, and vibe coders who want LLM-assisted development to be repeatable instead of chaotic.

Status

Live

Timeline

Released as an open-source workflow system

Domain

Developer education

Why

Open-source infrastructure

AI development workflow

Spec → Build → PR → Review

Live

Files

6

Examples

8

Setup

5m

A repeatable Claude Code workflow for solo builders, teams, and vibe coders.

Stack

Languages, services, data sources, and operating pieces behind the build.

MarkdownShellClaude CodeTemplatesExamples

Code Proof

What The Build Actually Contains

LOC

1.9k+

Files

16

Templates

Spec/PR

Setup

Shell

Workflow preview

Spec To PR

01

CLAUDE.md

Repo rules and build conventions

02

AGENTS.md

Agent autonomy boundaries

03

SPEC_TEMPLATE.md

Scope before code

04

PR_TEMPLATE.md

Explain the change

05

WORKFLOW.md

Repeatable loop

06

examples/

Real project starting points

Product proof

AI development workflow

Spec → Build → PR → Review

Live

Files

6

Examples

8

Setup

5m

A repeatable Claude Code workflow for solo builders, teams, and vibe coders.

Implementation

Code Behind The Surface

No code before a spec

md

The first rule is deliberately simple because it prevents the most expensive AI-coding mistake: building the wrong thing quickly.

## The Three Rules

**No code before a spec.**
The most expensive mistake in software is building the wrong thing correctly.
Five minutes of planning prevents hours of rework.

**No PR without a description.**
A PR with no description is a black box.

**No merge without a real test.**
"Works on my machine" is not a test.

Agent boundaries

md

The workflow teaches builders to separate safe autonomous actions from decisions that need human judgment.

## Always Stop and Ask a Human

- Deleting any file
- Changing the database schema
- Modifying authentication or authorization logic
- Changing payments or billing
- Changing production configuration

When in doubt, stop and ask.

Example-driven setup

sh

The setup path is designed to get a repo-specific CLAUDE.md in place quickly, then let builders tune it over time.

git clone https://github.com/dantrapp/claude-code-workflow.git
cd claude-code-workflow
chmod +x setup.sh
./setup.sh

Project Logic

Why This Exists

The point is not to show another screen. It is to show the gap, the build constraint, and the proof of work.

Mission

What would help people building with Claude Code understand the proper workflow before they accidentally create a pile of unreviewed AI-generated code?

Claude Code can write good code, but consistency depends on context and rules. A lot of new AI builders skip the spec, skip repo guidance, skip PR explanation, and skip real testing because nobody gave them a simple operating system.

Build

What Had To Work

I built a complete workflow kit: CLAUDE.md for project rules, AGENTS.md for autonomy boundaries, SPEC_TEMPLATE.md before code, PR_TEMPLATE.md after code, WORKFLOW.md for the whole loop, setup.sh for onboarding, and examples for common project types.

Why It Matters

Spec → Build → PR → Review

Gives AI builders a repeatable workflow before fast LLM output turns into unreviewed code debt.

Hard Parts

Teaching workflow without sounding academic

The project has to be simple enough for new builders and serious enough for experienced engineers. The language is direct: spec first, build inside rules, explain the PR, test for real.

Making structure feel lightweight

Too much process kills momentum. The kit keeps the core loop small, then uses examples for people who want stronger defaults.

Helping people avoid invisible AI debt

LLMs can create a lot of code quickly. The workflow slows down the risky parts: unclear scope, missing conventions, undocumented changes, and untested merges.

Decisions

Lead with a one-page workflow before deeper templates.
Make CLAUDE.md the repo-level source of truth for coding rules.
Use AGENTS.md to separate agent autonomy from human decisions.
Include examples for real project types instead of only abstract advice.

Next Move

I would add more examples, short walkthrough videos, and real before/after workflows showing how the system changes Claude Code output quality.

Tell Me About Your Project

Bring Me The Bottleneck.
I’ll Build The Answer.

Tell me what people are trying to do, where the current path breaks, and what kind of useful answer should exist.

Market Gap

Demand exists, but the answer is missing.

Workflow Drag

The work is still too manual, slow, or scattered.

Product Wedge

A small surface could prove the larger opportunity.

Quick Note