Molcajete.ai
Consistent, opinionated workflows for Claude Code. No AI multiple-personality disorder.
You know when you start using AI for your daily workflows, one day you get the right personality out of AI, other days you feel like it's the dumbest model in the world. What Molcajete.ai brings is the right setup for Claude Code — well-curated flows and consistent output. A curated marketplace of plugins that grind through complexity and blend workflows into smooth, production-ready solutions. Just opinionated commands and skills that get the job done, the way I like things done.
What is Molcajete.ai?
Molcajete.ai is a Claude Code plugin marketplace. It ships two plugins:
m — The core plugin. Generic, stack-aware commands covering the full development lifecycle, backed by a library of 18 reusable skills.legacy — A migration helper for moving files from the old v1 .molcajete/ format into the current m plugin format.
Installation
Prerequisites
Quick Install
/plugin marketplace add MolcajeteAI/plugin
Alternative: settings.json
For persistent configuration across sessions, add to .claude/settings.json:
{
"extraKnownMarketplaces": {
"molcajete": {
"source": {
"source": "github",
"repo": "MolcajeteAI/plugin"
}
}
}
}
Verify
/plugin marketplace list
You should see molcajete in the list. Both plugins are enabled by default:
{
"enabledPlugins": {
"m@molcajete": true
}
}
The m Plugin
The m plugin is the core of Molcajete.ai. Every command is a structured prompt that reads relevant skills, follows established conventions, and delegates complex sub-tasks to specialized agents.
Commands
| Command | Description |
|---|
/m:init | Initialize a new project with mission, tech stack, roadmap, and changelog |
/m:feature | Scope a new feature into structured requirements |
/m:refactor | Research a refactoring/replacement, produce impact analysis and requirements |
/m:spec | Create a technical specification for a feature or refactor |
/m:tasks | Break a specification into sequenced implementation tasks |
/m:dev | Implement a task from the task plan |
/m:fix | Diagnose and fix a bug or failed implementation |
/m:test | Write, run, or analyze tests for code |
/m:review | Code review on staged or recent changes |
/m:doc | Generate or update documentation for a file or directory |
/m:explain | Explain a topic in a clear, friendly 3-5 minute read |
/m:research | Deep research with tech stack context, parallel agents, and long-form output |
/m:copy | Rewrite text or file content with clear structure and good copy |
/m:prompt | Write a well-structured LLM prompt from a rough description |
/m:debug | Guided debugging workflow with hypothesis testing |
/m:commit | Create a well-formatted commit from staged changes |
/m:amend | Amend the last commit with staged changes |
/m:rebase | Interactive rebase helper |
Natural Language Triggers
Some skills activate automatically based on what you say — no slash command needed. The research-methods skill is the first to support this:
| What you say | What runs | Output |
|---|
| "What is a CLOB?" | Quick answer | Inline response, no agents |
| "Explain how OAuth works" | Explain flow (2 agents) | 3-5 minute Introduction, save prompt |
| "Research sharding for our Postgres" | Deep research (4 agents) | Full Learning Guide, save prompt |
The /m:explain and /m:research commands are shortcuts that skip classification and go straight to their respective depth.
Skills
Skills are reusable knowledge documents loaded by commands at runtime. Each skill encodes conventions, patterns, and standards for a specific domain.
