Skip to content
ML2

MCode Stack: MiniMax's coding-agent skills, open-sourced

Added Copyable prompt below

Published Last verified Released

Our verdict

Useful because it targets the habits around the agent rather than the prompt: when to hand work to a subagent, how to split a big change into small PRs, how long to wait on a long command, when a session has gone stale, and how to review a diff from a clean context. Each skill is plain Markdown you can read in a minute and copy one at a time, so the cost of trying it is low. It was published on 2026-10-28; the claim that the team uses it daily is MiniMax's own, and we haven't seen independent results yet, so start with one or two skills rather than the whole stack.

Who it's for
Developers who run Claude Code, Codex or MiniMax Code on long tasks and want written working rules for delegation, review and context instead of ad-hoc prompting
Leverage
A lab team's engineering checklists as ready-to-load skills, adopted one at a time
Effort
Low
Date added

What it does

MCode Stack (MiniMax-AI/mcode-stack, MIT) is a plugin marketplace with two stacks of SKILL.md skills. stacks/ronny holds 15: subagent-split-check (subagent or main loop), engineering-small-prs, engineering-change-descriptions, engineering-code-review, engineering-review-comments, engineering-review-feedback, engineering-emergency-changes, tool-budget-audit (what tools cost in context), context-tax-audit (what gets injected every request and how it affects the prompt cache), long-session-check, handoff-doc, long-command-check (a ten-point audit of long shell commands), independent-review (review a diff with only the diff, requirement and test results), harness-experiment and trajectory-review. stacks/jiachen adds lifting-loops and making-content. Each skill pairs with an article in MiniMax's MCode Engineering Notes. The skills use the common SKILL.md format, so most also work in other agents that read skills.

How builders use it

  • Install the main stack in Claude Code with claude plugin install ronny --marketplace MiniMax-AI/mcode-stack, or in Codex with codex plugin marketplace add MiniMax-AI/mcode-stack then codex plugin add ronny@mcode-stack.
  • Or copy a single skill folder, such as long-command-check, into ~/.agents/skills/ and open a new session.
  • Run independent-review on a finished diff before you merge, so a clean context checks the change against the requirement and test results.
  • Use subagent-split-check before fanning work out to parallel agents, and handoff-doc when a session gets long.

What to expect

Fewer stalled long-running commands, smaller reviewable PRs and diffs reviewed by a context that didn't write them, using rules you can read and edit.

Copy this: Install (Claude Code / Codex / one skill)

# Official source: github.com/MiniMax-AI/mcode-stack (MIT)

# Claude Code
claude plugin install ronny --marketplace MiniMax-AI/mcode-stack

# Codex
codex plugin marketplace add MiniMax-AI/mcode-stack
codex plugin add ronny@mcode-stack

# One skill for any agent that reads ~/.agents/skills
git clone https://github.com/MiniMax-AI/mcode-stack.git ~/mcode-stack
mkdir -p ~/.agents/skills
cp -R ~/mcode-stack/stacks/ronny/skills/independent-review ~/.agents/skills/

# Then ask your agent:
Run independent-review on the current diff against the requirement in docs/task.md and the latest test results.

Related discoveries

See the newest picks →
  • ToolCoding & Shipping Agents

    Elephant-Goldfish Model: specs reviewed by fresh agents

    Google's Apache-2.0 plugin that makes your coding agent draft a design spec with you, then has fresh zero-context subagents attack it before any code is written.

    Effort: MediumPrompt#Plugin / extension#Open source

  • ToolCoding & Shipping Agents

    Cloudflare security-audit skill: agent bug hunts

    Cloudflare's MIT skill that turns Claude Code, Codex or Cursor into a six-phase security auditor where a fresh agent tries to disprove every finding.

    Effort: MediumPrompt#Plugin / extension#Open source

  • ToolCoding & Shipping Agents

    claude-mem: memory across sessions for coding agents

    An Apache-2.0 plugin that watches your coding agent's sessions, compresses what it did into searchable notes, and injects the relevant ones into the next session.

    Effort: LowPrompt#Plugin / extension#MCP

  • ToolCoding & Shipping Agents

    improve: shadcn's skill where a strong model plans

    An agent skill (MIT) from shadcn, the creator of shadcn/ui, that uses your most capable model to audit a codebase and write self-contained implementation plans, then hands those plans to cheaper models to carry out. It never edits your source code itself.

    NewEffort: LowPrompt#Plugin / extension#Open source

Get the best new A.I. picks in your inbox

The best new agent workflows and tools we’ve vetted, with a clear verdict on each. Short, occasional, useful.

More in Coding & Shipping Agents →

Free. No spam. Unsubscribe anytime.