We deploy a walking skeleton before we write a feature

The demo that works on one laptop

Most projects start with feature work. Someone picks the highest value screen, builds it locally, and by the end of week one there is something to show. It feels like progress. Then integration, deployment, auth, environments and real data all arrive at once, usually late, and the schedule that looked comfortable stops being comfortable.

We do the opposite. Before any feature, we build and deploy a walking skeleton: a thin slice that runs the whole path from client to database and back, in the real production environment, doing the smallest possible real thing.

What the skeleton actually contains

• A deployed frontend that renders one real screen
• An API it calls over the network, not a mock
• One endpoint that reads and writes one table
• Authentication on that endpoint, even if there is one user
• The pipeline that built and shipped all of it, triggered by a merge to main

It is deliberately boring. The screen might show a single record. The point is not what it does, it is that every seam between the parts exists, is wired up, and is running where customers will eventually hit it.

Why it pays for itself

The hard parts of software are rarely the features. They are the joins: the deploy that only fails in staging, the CORS rule nobody owns, the migration that locks a table, the secret that is missing in one environment. A walking skeleton forces every one of those problems into week one, when there is slack to solve them and no feature work is blocked.

It also changes how the team works for the rest of the project. Once main deploys to production automatically and safely, shipping is a non event. Features go out in small pieces behind flags. Nobody batches a month of work into a release and holds their breath.

What it costs

Two to five days, usually, depending on how much infrastructure already exists. On a greenfield build it is the first thing we do. On an existing system it is often already there in some form, and the exercise becomes making the deploy path fast and trustworthy.

That is a real slice of a short project. It is also the cheapest insurance you will buy, because the alternative is discovering the same problems in the last two weeks instead of the first.