joshft

joshft / correctless

Public

Correctness-oriented development workflow skills for Claude Code

24
1
89% credibility
Found Mar 29, 2026 at 24 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Shell
AI Summary

Correctless is a plugin system for Claude Code that guides AI-assisted coding through structured phases of planning, testing, reviewing, and verifying using multiple AI perspectives to produce more reliable code.

How It Works

1
💡 Discover Correctless

You hear about Correctless, a helpful guide that trains your AI coding buddy to plan carefully, test thoroughly, and double-check work to avoid sloppy mistakes.

2
📦 Add the guide to your AI tool

In your AI coding helper called Claude Code, you grab Correctless from the easy plugin store and pick the lite version for everyday projects or full for important ones.

3
🛠️ Prepare your project

With one simple command, it scans your work area, sets up safety rules, and creates note templates so everything stays organized from the start.

4
🌿 Start a new feature branch

You make a fresh branch for your new idea and kick off by chatting with the AI to define clear rules for what success looks like.

5
🔍 Review the plan skeptically

A fresh AI view scans your plan for gaps, edge cases, and past mistakes, suggesting fixes before any building begins.

6
🧪 Test first, then build

The AI writes tests based on your plan that must fail at first; then it builds code to make them pass, with locks preventing shortcuts.

7
👀 Quality check and verify

Separate AI checkers hunt for bugs, confirm the code matches the plan perfectly, and ensure no sneaky issues slipped in.

🎉 Ready to share your solid work

Your feature is bug-free, well-documented, and ready to blend into the main project, giving you confidence it won't break later.

Sign up to see the full architecture

6 more

Sign Up Free

Star Growth

See how this repo grew from 24 to 24 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is correctless?

Correctless brings correctness-oriented development to Claude Code via shell-based skills and hooks. It enforces a strict workflow—spec before code, tests before implementation—using separate Claude agents for spec writing, testing, coding, QA, and verification to avoid self-grading biases that plague AI-assisted coding. Developers get slash commands like /cspec, /ctdd, and /cverify that gate edits by phase, ensuring you can't code without a reviewed spec or QA without passing tests.

Why is it gaining traction?

Unlike plain Claude Code sessions where agents rush happy-path code and overlook edges, Correctless structures prompts for adversarial reviews and mutation testing, catching bugs early. Lite mode adds ~5 minutes per feature for everyday apps; Full ramps to 15-30 minutes with formal Alloy models and STRIDE threats for high-stakes work. The phase-locking hooks make sloppy workflows impossible, hooking devs tired of production surprises.

Who should use this?

Backend engineers building APIs or CLIs in Go, TypeScript, Python, or Rust, where test runners exist. Security devs on proxies or auth systems needing Full mode's audits. Solo devs or small teams using Claude Code CLI who want enforced TDD without discipline lapses.

Verdict

Grab Lite for routine features if you're deep in Claude Code—solid docs and test coverage make setup painless despite 24 stars and 0.9% credibility score signaling early days. Skip Full unless bugs cost real money; it's overkill for prototypes.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.