Ship: Team Task Manager
The Level 2 finale - specify, build, verify and deploy a real multi-user product with roles.
Before this: Testing Your App
Objective
Combine every Level 2 skill into one delivery: write the brief first, direct the full build of a multi-user task manager with roles and guarded APIs, keep the suite green, and deploy it publicly with demo accounts.
Vocabulary
The words to use when directing your agent.
- feature spec
- The brief document describing what to build before building - purpose, features, acceptance criteria, non-goals.
- scope discipline
- Cutting features to protect quality gates - deciding what NOT to build.
- smoke test
- A quick post-deploy pass over critical flows proving the deployment breathes.
- demo account
- A throwaway seeded account letting outsiders try the app safely.
Understand it
Composition is the exam
Nothing in this project is new: servers, schemas, validation, authN, authZ, tests, deploys - every ingredient exists in your earlier builds. What IS new is holding all of them in one coherent specification and verifying them as a system. That integration burden is the actual skill being certified.
Scope cuts are wins
Due dates, notifications and comments all sound vital until the suite is red and guards are unaudited. Cutting them - visibly, in writing, as scope decisions - IS the senior move this module teaches.
Core concepts
The brief comes first - always
You wrote a briefing document in Level 0; now it carries real weight. Features, acceptance criteria and non-goals agreed BEFORE code means every later argument is settled by pointing at the doc, not by renegotiating mid-build with an eager agent.
Quality gates outrank features
When schedule pressure meets quality gates, features lose. That ordering - suite green, guards audited, deploy verified - is what separates professional shipping from hope-driven deployment. A smaller shipped thing that works beats a bigger broken thing.
Examples
A right-sized v1
Workspaces with two roles; tasks with title, assignee and three statuses; membership managed by owners. Everything else - labels, search, activity feeds - explicitly listed as non-goals. That fits in days and ships solid.
Demo accounts done right
Seed demo-owner@ and demo-member@ accounts with obviously fake data and throwaway passwords, documented in the README. Outsiders explore freely; nothing real is at risk. Never seed real-looking personal data for demos.
Common mistakes
Common mistake
Letting the agent talk you into extra features mid-build.
Fix
New ideas go on a parking-lot list in the brief. v1 ships the agreed slice; the parking lot waits for v2.
Common mistake
Deploying with the suite red because 'the feature works locally'.
Fix
Red suite equals no deploy. Fix, revert or cut - those are the only doors out.
Working with your AI agent
Structure the whole engagement as phases mirroring your skills: brief approval, schema and permission matrix approval, API skeleton, authN+authZ, test suite green, deploy rehearsal, production smoke. Demand a short completion report per phase against ITS exit criteria - you are running the loop you learned, at project scale.
Build it
The module’s real project. Your agent builds; you direct and inspect.
Project: Team Task Manager, Shipped
The Level 2 capstone: a deployed multi-user task manager - workspaces with owner/member roles, assignable tasks with a status flow - guarded, tested, verified in production, with demo accounts for visitors.
Requirements
- Brief document approved BEFORE implementation (purpose, features, acceptance criteria, explicit non-goals)
- Workspaces: owners manage members; members act only within their workspaces
- Tasks: title, optional description, assignee among members, status flow todo-doing-done
- Full security stack active: hashed passwords, session cookies, permission matrix, whitelist validation, uniform errors
- Test suite (10+) green pre-deploy covering auth journey, role boundaries and task flows
- Deployed publicly with two documented demo accounts and post-deploy smoke test executed
- Rollback rehearsed once on the real deployment
Context for your agent
Run this build as phased delivery with me approving gates: phase 1 the brief (challenge my scope, propose cuts), phase 2 schema plus permission matrix, phase 3 API skeleton with validation, phase 4 authN/authZ with guards, phase 5 tests-first until green, phase 6 deploy rehearsal plus production smoke. Hold ME to the gates too - if I skip one, refuse to proceed. Non-goals I am declaring upfront: no comments, no notifications, no due dates, no search.
Verify it - twice
Check every item yourself first. Then hand your agent the prompt and compare verdicts.
Verification prompt - paste into your AI agent
Conduct the final acceptance review of my Team Task Manager as the release gatekeeper. Work through: (1) does the shipped build match the approved brief - name any undocumented deviation, (2) execute the full security stack checklist yourself (hashes, cookies, generic errors, validation, envelope, log hygiene), (3) run the test suite and the production smoke script, reporting results honestly, (4) attempt three cross-role exploits of your own invention and report outcomes, (5) verify demo accounts work as documented and leak nothing real. Finish with a SHIP or DO-NOT-SHIP verdict and, if the latter, the exact blocking list.
Run this against “Ship: Team Task Manager” after your own checklist pass. Where the agent and you disagree, investigate before moving on.
Review questions
Answer out loud. If you cannot, revisit before moving on.
Which feature did you CUT, and how did you document that decision?
Replay your six build phases and the gate you enforced at each one.
What would you do differently if this were ten workspaces instead of two users?
You can now
- Deliver a complete multi-user product through gated phases
- Hold scope discipline under an eager agent's enthusiasm
- Gate releases on evidence - suites, smokes and audits - not vibes
Up next
Caching Layers