Your first sprint

From idea to shipped code in minutes. The agent does the work. You guide the scope.

Configure your project

If you haven't already, run the setup conversation. It asks about your project type, detects your stack, and sets preferences.

/nano-run

You only need to do this once per project. After that, every skill knows your stack, your conventions, and your preferences.

Describe what you want

Tell your agent what to build. Be specific about the outcome, loose about the implementation. For this walkthrough, we will build a simple habit tracker.

> Build a habit tracker. Users can add habits, check them off daily,
  and see a streak count. Use SQLite for storage. No auth for now.

Run the full sprint

/think --autopilot

This kicks off the entire sprint pipeline. Each phase runs automatically, and the agent moves to the next one when it finishes. Here is what happens:

/think challenges your scope

Before writing any code, /think pushes back. It asks whether you really need SQLite or if a JSON file would be simpler for a v1. It flags that "streak count" has edge cases around timezone handling. It suggests starting with three habits max to keep the first version tight.

/nano plans the implementation

Once scope is locked, /nano produces an implementation plan. For our habit tracker, it plans 6 files: a schema, a migration script, three route handlers, and a minimal frontend.

Plan: 6 files, 3 routes, 1 migration
├── db/schema.sql
├── db/migrate.sh
├── src/routes/habits.ts
├── src/routes/checkins.ts
├── src/routes/streaks.ts
└── src/ui/index.html

Build creates the files

The agent writes all 6 files following the plan. No hallucinated extras, no missing pieces. The plan is the contract.

/review finds issues

Two-pass review. Structural pass catches that the migration script is missing a IF NOT EXISTS guard. Adversarial pass finds that the streak calculation breaks if you check in twice on the same day. The agent auto-fixes the first issue and flags the second for your decision.

Review: 2 issues found
  ✓ Fixed: migration missing IF NOT EXISTS
  ⚠ Needs input: duplicate check-in on same day

/security grades the code

Runs OWASP Top 10 checks and STRIDE threat modeling against your code. Our habit tracker gets a grade A because there is no auth surface, no user input that reaches SQL without parameterization, and no external API calls.

/qa runs tests

Generates and runs tests based on the implementation plan. For our tracker: 24 tests, 24 passing. Covers happy paths, edge cases from the review phase, and the streak calculation logic.

QA: 24/24 passing
  12 unit · 8 integration · 4 edge cases

/ship commits and opens a PR

Creates a commit with a structured message, opens a PR with the sprint summary linked, and suggests what to build next.

What you end up with

• Working code: 6 files, tested and reviewed
• 6 artifacts: one per phase, each a structured JSON record of what happened
• A sprint journal: the full decision trail from scope to ship
• Next feature suggestions: based on what /think flagged as out of scope

The artifacts are not just logs. They are structured data that future sprints can read. When you run /feature to add daily reminders next week, the agent already knows your schema, your conventions, and what you deliberately left out of v1.