Skip to main content

The Lifecycle of one Work-item

Purpose: The 7-step lifecycle every work item moves through from author to close — and the gates that protect description quality.

Reading time: ~3 minutes · Audience: BA / analyst, tech lead, QA · The composition principles are in Work-Item Description Composition; the per-type templates are in Templates for Work-items of different types.

The lifecycle is the same regardless of WI type — what differs is which agents do which steps. Knowing the lifecycle stops teams from skipping the validation gates that protect description quality.

The seven steps

#StepOwnerWhat happens
1BA writes the WIBA / analystAuthor per the type template — see Work-Item Type Templates. Include scope-out, sibling reference, mapping table where applicable.
2Tech lead validatesTech lead / architectIs scope atomic? Are dependencies explicit? Is the contract there? Reject and send back if any of these fail — do not push half-ready WIs into agent runs.
3Swifter generates the specificationAnalyst agentFor component / page → detailed spec; for AC → criteria. The spec is the artefact the developer agent will read.
4BA compares the generated spec to expectationBA / analyst
Development Team
~90 % match → ship to development. Less → refine the WI and regenerate. The 90 % gate is the cheapest place in the lifecycle to fix authoring problems.
Development Team representatives can be also involved at this Step if necessary to validate details of low-level specification.
5Swifter executes the WI against the validated specDeveloper agentImplementation step. Review in the Dev / App tabs as work lands.
6QA runs Playwright tests against ACQA Tester agent + QA ownerGenerates test cases and scripts; runs against Preview; iterates on failures.
7CloseOwner of the parent featureMark the WI Done; flip the parent feature's child entry; merge or note any rework decisions.

Why every gate matters

If the lifecycle is shortened — for example, the BA skips step 2's validation, or the team skips the spec comparison at step 4 — the description-quality problems land in the developer agent's input and cost a session to unwind. The discipline is procedural and cheap.

The two gates that catch problems early:

  • Step 2 — Tech lead validates catches scope and dependency problems before the analyst burns a session on a half-ready WI.
  • Step 4 — BA compares catches misalignment between expectation and generated spec before the developer agent runs. A wrong spec produces wrong code; catching it here costs minutes, catching it after implementation costs a session.

What comes next

The per-type agent + command sequences for steps 3 and 5 are in Work-Item Run Sequences. The diagnostic guide for when a step stalls is in Cookbook → Journeys → Troubleshooting.