ChengRang

MiniMax MCode Stack

AI Coding Open Source

MiniMax's MIT-licensed agent skill library released 2026/10/28, holding 17 SKILL.md files and five checklists across two stacks that capture the team's daily workflows

MiniMaxopen sourceagent skillsworkflowMIT
Visit MiniMax MCode Stack

Disclaimer: Review content represents our editorial team's views and experience, not commercial recommendation or investment advice. Product info and pricing may change; refer to official sources.

Overview

MCode Stack is an open-source library of agent skills that MiniMax released under the MIT licence on October 28, 2026. It contains the workflows the MiniMax Code team uses every day to build MiniMax Code. The repository is explicit that this is not a prompt collection or a list of best practices.

There are two stacks today, ronny and jiachen, holding 17 SKILL.md files and five checklists. Ronny's stack, from the R and D lead, covers task splitting, context cost, long-session hygiene and independent review. Jiachen's two skills find the loops where a person is acting as a manual step for agents and turn each into a fix, and turn source material into Xiaohongshu image posts and vertical video.

There are three ways to install: copy a single skill directory into ~/.agents/skills, copy a whole stack as a local plugin, or add the repository as a plugin marketplace and install, update or uninstall from there. Because the format is the common SKILL.md, most of the skills also work in Claude Code.

Key Features

Use Cases

Pros

Pricing

MIT licensed and free to use or modify. The six engineering series skills adapt material from Google Engineering Practices and remain under CC-BY 3.0, with terms in stacks/ronny/NOTICE.md. Cloning the repository is all that is required; there are no server costs.

Summary

What matters here is not the skill count but the record of how a company building a coding agent constrains its agents day to day. Splitting work, controlling context, managing long sessions and reviewing independently are habits everyone nods at in an article and almost nobody executes. Turning them into installable skills is what makes them run. Start with independent-review and long-command-check, where the difference shows up immediately.

Version History

Category
AI Coding
Pricing
Open Source
Tags
MiniMax · open source · agent skills
Website

Related Tools