gitwtfhub

wtf is nucleus-marketplace?

protect-my-hair/nucleus-marketplace — explained in plain English

Analysis updated 2026-05-18

43PythonAudience · developerComplexity · 4/5LicenseSetup · moderate

TL;DR

A Claude Code and Codex plugin that enforces structured, auditable, and traceable output at every step of AI assisted coding.

Mindmap

mindmap
  root((Nucleus))
    What it does
      Adds discipline to AI coding workflows
      Enforces structured skill outputs
      Blocks writes outside approved boundaries
    Tech stack
      Python
      Claude Code
      Codex
    Use cases
      Audit AI generated code changes
      Gate writes to sensitive paths
      Produce traceable delivery evidence
    Audience
      Development teams with code review
      AI coding agent users

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

Why would anyone build with this?

REASON 1

Add structured, auditable checkpoints to an AI coding agent's workflow.

REASON 2

Enforce that AI generated file writes stay within approved directories.

REASON 3

Require an independent sub agent review before any human review request is sent.

REASON 4

Generate a traceable evidence package, including a pull request draft, at the end of a feature.

What's in the stack?

PythonClaude CodeCodex

How it stacks up

protect-my-hair/nucleus-marketplacealibaba/omnidoc-tokenbencharccalc/dwmfix
Stars434343
LanguagePythonPythonPython
Setup difficultymoderatemoderateeasy
Complexity4/53/52/5
Audiencedeveloperresearchergeneral

Figures from each repo's GitHub metadata at analysis time.

How do you spin it up?

Difficulty · moderate Time to first run · 30min

Requires an AGENTS.md file at the repo root and an already initialized git repository before it will run.

Use freely for any purpose, including commercial use, as long as you keep the copyright notice.

Wtf does this do

Nucleus is a plugin for AI coding agents such as Claude Code and Codex. Rather than helping the AI write more code, it adds discipline around how that code gets produced, so every step the AI takes leaves behind output that can be checked for structure, kept within defined boundaries, and traced back to a decision. The README explains the problem it targets: it is not that AI writes wrong code, but that once it finishes, nobody can easily tell what it actually did, what it based its decisions on, or whether the result should be trusted. Common issues it names include starting to code before requirements are clear, losing track of which version of a change was actually confirmed by a human, an AI reviewing its own work in the same chain that produced it, and decisions that only exist in a chat history that eventually gets summarized away and lost. To address this, each unit of work Nucleus manages, called a Skill, has a defined input and output format described as a JSON schema. Rules about which files and folders can be written to are checked automatically while the agent runs, and a violation blocks the action rather than being silently allowed. Before any request for human review goes out, an independent sub agent runs a pre-review check first. If something fails, the tool's stated policy is to alert and block rather than quietly downgrading or skipping the check. Nucleus provides 26 separate Skills covering the development lifecycle, from analyzing a requirement and building a plan, through implementing a feature with multiple review passes, to generating and running tests, and finally assembling delivery evidence such as a pull request draft. It requires an AGENTS.md file at the root of the target code repository, which it treats as the primary source of truth, and the target repository must already be a git repository. It installs into Claude Code or Codex through a plugin marketplace command, and is intended for teams that already have a code review process and want AI generated work to plug into it with a clear trail of evidence.

Yoink these prompts

Prompt 1
Help me install the Nucleus plugin marketplace for Claude Code and add the Nucleus skill.
Prompt 2
Walk me through Nucleus's workflow from requirement analysis to delivery evidence for a new feature.
Prompt 3
Explain what an AGENTS.md file needs to contain for Nucleus to run in my repository.
Prompt 4
Show me how Nucleus's asset boundary check blocks writes to legacy or restricted paths.

Frequently asked questions

wtf is nucleus-marketplace?

A Claude Code and Codex plugin that enforces structured, auditable, and traceable output at every step of AI assisted coding.

What language is nucleus-marketplace written in?

Mainly Python. The stack also includes Python, Claude Code, Codex.

What license does nucleus-marketplace use?

Use freely for any purpose, including commercial use, as long as you keep the copyright notice.

How hard is nucleus-marketplace to set up?

Setup difficulty is rated moderate, with roughly 30min to a first successful run.

Who is nucleus-marketplace for?

Mainly developer.

View the repo → Decode another repo

This repo across BitVibe Labs

Don't trust strangers blindly. Verify against the repo.