mahu 1.1.0

Cross-agent /mahu entrypoint for context, prototypes, presentations, tests, feedback loops, and growth

Mahu

I'm Mahu, your free AI work buddy for context, prototypes, tests, feedback, and growth.
Bring me into your daily workflow.

Mahu is not a single builder and not another chat persona. It is a cross-agent /mahu entrypoint that helps Codex, Claude Code, OpenCode, Copilot, Trae, WorkBuddy, and future agents work through the right SOP instead of improvising every time.

Use /mahu when daily work needs structure: capture context, build a prototype, create a presentation, test the result, publish it for feedback, or turn the learning loop into growth momentum.

What Mahu Helps With

User intent	Mahu subskill	Owning tool
context, memory, topics, requirements	context	fcontext
prototype, UI, website, app, design system	prototype	fdesign
deck, PPT, slides, presentation	presentation	fppt
review, feedback, comments, upload, resolve	review	floop-client
tests, QA, regression, acceptance	test	testboat
growth, iteration, learning loop	combine context, test, and feedback	Mahu SOP

Quick Start

Ask your agent to install Mahu:

Install skill for me: https://github.com/lijma/mahu

Then use Mahu from the agent chat:

/mahu create a product prototype for a checkout flow
/mahu build a deck about how to create an agent skill
/mahu upload this version for review and collect feedback
/mahu save this decision as project context
/mahu run a smoke test and summarize the risk

That is the intended flow: give the repo to your agent, let it inspect the Mahu adapter for the current environment, then invoke Mahu with /mahu.

Install

Most users should use the Quick Start above. Mahu includes adapters for Codex, Claude Code, OpenCode, GitHub Copilot, Trae, and WorkBuddy, so the agent can choose the correct install shape after reading this repository.

If you want to be explicit, say:

Install skill for me: https://github.com/lijma/mahu
After installation, I want to use Mahu by typing /mahu.

The agent should inspect the repo, pick the adapter for itself, install the right files, and verify the setup before using Mahu.

How It Works

When /mahu is invoked, the agent should:

1. Load Mahu through the installed adapter.
2. Read the bundled SKILL.md.
3. Choose the right subskill with AI judgment.
4. Load only the needed file under skills/.
5. Run mahu doctor --subskill <name> or check the dependency directly.
6. Follow that subskill's validation loop.

Each Mahu subskill declares its own dependency:

Subskill	Required CLI	Typical install
context	fcontext	pip install fcontext
prototype	fdesign	pip install fdesign
presentation	fppt	pip install fppt
review	floop	pip install floop
test	testboat	pip install testboat

Advanced

The CLI is optional. It exists so agents can perform deterministic install, validation, and dependency checks when a manual path is useful.

Install from PyPI:

pip install mahu

Or install from GitHub:

pip install git+https://github.com/lijma/mahu.git

Install Mahu into a specific agent workspace:

mahu enable claude --target .
mahu enable codex --target .
mahu enable copilot --target .
mahu enable opencode --target .
mahu enable trae --target .
mahu enable workbuddy --target .

What mahu enable writes:

Agent	Files created
Claude Code	.claude/plugins/mahu/ with .claude-plugin/plugin.json and bundled skills/mahu/
Codex	.codex/skills/mahu/
GitHub Copilot	.github/skills/mahu/ and .github/instructions/mahu.instructions.md
OpenCode	.opencode/skills/mahu/, .opencode/commands/mahu.md, and a Mahu section in AGENTS.md
Trae	.trae/skills/mahu/ and .trae/rules/mahu/rule.md
WorkBuddy	.workbuddy/skills/mahu/

Validate the repository and dependencies:

mahu validate
mahu doctor --subskill prototype

Development

PYTHONPATH=src pytest --cov=mahu --cov-report=term-missing --cov-fail-under=100