Sprint planning & autosprint

A sprint in SDLC Studio is a batch of delivery work driven to a goal, opened deliberately and closed with ceremony. Planning is where the batch is chosen, sized against reality, and gated before any code is written.

The breakdown gate

sprint plan refuses a batch that is not ready: an un-refined request, an un-sized unit, a story whose acceptance criteria are still placeholders. The refusal is the point - it stops a plan filling with work nobody scoped.

/sdlc-studio sprint plan --stories Ready --epic EP0001 --goal done --write

--write opens the run: it stamps a run id, the start time, the approved batch and a token baseline, and it archives any previous run. Before that, plan runs a git fetch and an origin-drift pre-flight, so you never plan against a stale checkout.

Goal ladder

A plan targets a rung: triage -> plan -> design -> done. Run a single rung for a checkpoint (groom the batch, produce a design), or --goal done to take it all the way. The rung is recorded on the run, so the close can say what the sprint was actually for.

Capacity and appetite

The plan measures the batch against a standing appetite (the sprint’s capacity in units and minutes) and records the accepted appetite it was planned at. If a batch exceeds the standing appetite, the overage is flagged and the run is reported against the standing figure - so a decision to take on more than fits is written down, never hidden by quietly raising the ceiling.

The test strategy

The plan derives, from your TSD, which risk areas the batch touches and which coverage the TSD demands that the batch would not deliver - so gaps are named at plan time, not discovered at review.

Autosprint - the goal-driven loop

An agent runs in a loop that cannot judge its own exit condition. The lifecycle has always been that loop: specify, build, validate against the specification, reconcile, repeat - with acceptance criteria as the test that closes it. sprint --goal done (autosprint) runs it: a prioritised batch, built and verified rung by rung, stopping when its acceptance criteria are met, closing with a reconcile and review.

The close

sprint close is one deterministic chain: goal verdict, retro validate and extract, lessons summary, the close gate (reconcile, review currency, the retro), a handoff if the run stopped short, and the review anchor. Past the two-role threshold, --apply-signoff fans an operator’s sign-off into per-unit Done transitions. See two-role review.