MarcosSanchez7020

Mm Method — AI skill for Claude Code

AI community

A working methodology for software projects built with AI coding agents (Claude Code, optionally with Codex): owner decisions as numbered lists, project memory in the repo, plan before coding, indepen.

How to install Mm Method

This entry records only its repository, not the path inside it, so there is no exact command to give. Open MarcosSanchez7020/mm-method and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Mm Method does

Alternatives in AI

Gentle AI — Agent Skills Index — When working on this project, load the relevant skill(s) BEFORE writing any code 1.9k ★
Session Kit — Close the terminal windows, not the AI coding sessions 649 ★
LLM Council Skill — A skill to claude code that enables brainstorming with other LLMs (ChatGPT, Gemini) before presenting the impl 416 ★

README

mm-method

**English** · [Español](README.es.md)

A working methodology, packaged as a Claude Code and Codex plugin, for software projects built with AI agents in which one person (the owner) makes the technical decisions. It works with Claude Code or Codex, alone or together in the same repository.

It is not a library or a tool that automates anything: it is a set of instructions that the agent follows. What makes it hold in a project is the project's `CLAUDE.md`, the protection of `main` and the team's discipline. The skill teaches the agent how to work and gives it the skeletons to set it up.

The problem it solves

A coding agent is fast and capable, but in a project that lasts weeks the same problems show up:

it makes technical decisions that were not up to it, or makes them silently;
what was decided in one chat does not reach the next one, or another computer, or the other agent;
it codes before it is clear what needs to be coded;
it writes tests that copy what the code returns, and so they always pass;
two agents in the same repo step on each other's branches and files;
context and weekly usage are spent without anyone noticing.

The methodology answers each one with a concrete rule and the reason for that rule.

The principles

#	Principle	Why
1	The technical source prevails (the standard, the specification)	If the product has to comply with something, no decision, neither the owner's nor the agent's, contradicts it silently.
2	Assumptions and limitations in plain sight	The product's user and the owner always know what was assumed and what is not checked.
3	The owner decides; the agent proposes	Decisions are requested in a numbered list, with the recommended option in bold, and are recorded verbatim.
4	First plan, then code	No product code is written without the owner's word.
5	Nothing made up	No data, results or capabilities; a test's expected values are not copied from the c