For Codex & Claude Code

More intent.
Less prompting.

Give your coding agent the workflows, context, and engineering judgment to move your ideas forward.

Get just-vibe Explore the commands

The just-vibe mark rendered as a sculptural lime-green path

A little structure.A lot of possibility.

218purposeful workflows

113engineering profiles

22specialist packs

Yours.MIT licensed. User owned.

Go from intention to action

Say what you need.
Keep your context.

Pick a workflow. Add the details that matter. Each command gives your agent a method, a scope, and a way to check its work.

A little context goes a long way

/just-vibe:auto

See how it works

A prompt builder. Your coding agent runs the workflow in your project.

Built for the work behind the work

From the first component
to the next model.

Browse all 221 commands

218 workflows + 3 convenient aliases. The catalog is generated from the shipped package.

A different lens for every task

Same agent.
Sharper perspective.

A frontend engineer checks interaction and state. An ML engineer checks splits and reproducibility. A principal engineer checks the boundaries.

Profiles give your agent relevant priorities and verification steps. You keep the decisions.

Find your profile

application

Frontend engineer

Trace component state, URL state and server state separately.

Machine learning

Machine learning engineer

Define prediction time, label horizon and valid splits.

technical leadership

Principal engineer

Identify systemic constraints, option value and irreversible decisions.

security

Application security engineer

Trace attacker-controlled input to privileged operations.

A toolkit with follow-through

The details
make the difference.

Read the field guide

Keep what you’ve learned.

Save decisions and corrections into project instructions with remember. Check for conflicts as the work changes.

Show the work behind “done.”

Workflows call for relevant verification and honest coverage limits. Proof organizes the evidence your agent actually collected.

Learn something along the way.

Use teach to understand a topic or a workflow’s implementation, with worked examples and optional practice.

One install. A new starting point.

Bring your next idea.

Install the toolkit, start a new conversation, and put it to work.

More intent.Less prompting.

Say what you need.Keep your context.

From the first componentto the next model.

Same agent.Sharper perspective.