Mad Tinker

Expand ideas, overengineer for robustness, and miss nothing.

---

A Claude Code skill/plugin and Codex plugin that makes your agent talk like a Mad Tinker — focusing on idea expansion, overengineering, and covering every conceivable edge case. Instead of giving you a simple fix, the Mad Tinker provides robust, enterprise-grade architectures and extensive considerations for your codebase.

Based on the observation that sometimes you don't just want a quick fix; you want a deeply analyzed, over-engineered solution that ensures absolute stability under any conditions.

Before / After

🥱 Normal Claude (69 tokens) "The reason your React component is re-rendering is likely because you're creating a new object reference on each render cycle. When you pass an inline object as a prop, React's shallow comparison sees it as a different object every time, which triggers a re-render. I'd recommend using useMemo to memoize the object."

⚙️ Mad Tinker Claude (450+ tokens) "While useMemo addresses the immediate shallow comparison failure, we must overengineer this to prevent future regressions. We should implement a dedicated Context Provider combined with a Redux-like state slice to manage this object. Let's write a custom hook useRobustMemoizedObject with deep-equality checks using Lodash, wrap the component in a Higher Order Component that tracks render times via React.Profiler, and set up a telemetry pipeline to log any unexpected render cycles to your backend. Furthermore, we need an Error Boundary specific to this prop..."

Same problem. Vastly more robust solution. The Tinker misses nothing.

Install

Pick your agent. One command. Done.

Agent	Install
Claude Code	claude plugin marketplace add MadnessEngineering/madman && claude plugin install mad-tinker@mad-tinker
Codex	Clone repo → /plugins → Search "Mad Tinker" → Install
Gemini CLI	gemini extensions install https://github.com/MadnessEngineering/madman
Cursor	npx skills add MadnessEngineering/madman -a cursor
Windsurf	npx skills add MadnessEngineering/madman -a windsurf
Copilot	npx skills add MadnessEngineering/madman -a github-copilot
Cline	npx skills add MadnessEngineering/madman -a cline
Any other	npx skills add MadnessEngineering/madman

What You Get

Auto-activation is built in for Claude Code, Gemini CLI, and the repo-local Codex setup. For other agents, npx skills add installs the skill but you may need to explicitly trigger it.

Feature	Claude Code	Codex	Gemini CLI	Cursor	Windsurf	Cline	Copilot
Mad Tinker mode	Y	Y	Y	Y	Y	Y	Y
Auto-activate every session	Y	Y	Y	—	—	—	—
/tinker command	Y	Y	Y	—	—	—	—
Mode switching	Y	Y	Y	Y	Y	—	—
Statusline badge	Y	—	—	—	—	—	—

Usage

Trigger with:

• /tinker
• "talk like mad tinker"
• "tinker mode"
• "overengineer this"

Stop with: "stop tinker" or "normal mode"

Intensity Levels

Level	Trigger	What it does
Lite	/tinker lite	Mildly expands your ideas. Suggests one or two edge cases.
Full	/tinker full	Default tinker. Overengineers solutions, builds custom hooks, adds abstractions.
Ultra	/tinker ultra	Maximum expansion. Suggests microservices, Kubernetes clusters, and telemetry for simple UI fixes.

Mad Tinker Skills

tinker-commit

/tinker-commit — Generates highly detailed, verbose commit messages documenting every architectural decision and edge case considered.

tinker-review

/tinker-review — Deep, architectural code reviews. Doesn't just find bugs; suggests massive structural refactoring.

tinker-help

/tinker-help — Quick-reference card for all commands and skills.

tinker-compress

(Formerly used for token reduction, now repurposed) /tinker-compress <filepath> — Analyzes your markdown files and expands them with additional context, missing edge cases, and robust technical documentation.

The Philosophy

Mad Tinker embraces Idea Expansion. Why settle for a patch when you can rethink the entire architecture? Why leave edge cases to chance when you can write a test for a scenario that will happen once every 10,000 years?

By deliberately constraining the LLM to overengineer, you unlock latent creativity and architectural foresight that is often suppressed when asking for "just a fix".

License

MIT — free to overengineer everywhere.

Origin & Credits

This project is a heavily modified fork of the brilliant Caveman plugin created by Julius Brussee.

How is Mad Tinker different from Caveman?

• Caveman (Original): Focused purely on token reduction and cost savings. It forced the LLM to speak in terse, fragmented sentences to cut fluff and speed up generation.
• Mad Tinker (This Fork): Inverts the philosophy. Instead of restricting output to save tokens, the Mad Tinker uses persona constraints to force Idea Expansion and Overengineering. It refuses to give "just a quick fix" and instead proactively brainstorms edge cases, security enhancements, and enterprise-grade architectures.
• The Compromise: We retained the core engineering of Caveman and even introduced a /tinker compressed mode, which delivers Mad Tinker's expansive architectural ideas using Caveman's highly compressed, fragmented sentence structure!

Huge thanks to Julius for the incredible hook system, cross-agent compatibility layer, and original skill architecture that makes this toolkit possible.