Planner

Planner turns "what does this release promise?" into a single artifact that stays attached to the answer through build, run, and sign-off. It reads a ticket, a doc, a recording, or a Slack thread and lays out what needs to be true as a list of acceptance criteria — then tracks, for each one, whether a test exists, whether it passed, and whether someone has actually signed off on it.

Everything in Planner hangs off one identifier: a criterion's id (AC-3). A test declares which criterion it proves, a run reports pass/fail per criterion, a shared report opens on criteria, and sign-off blocks on any criterion left unmet — not on the file that happens to exercise it.

The workflow

Planner is six screens, worked roughly in order — though nothing forces you through them linearly, and a plan can sit at any stage indefinitely.

1. Intake

Answering "what do we test?" starts from something that already exists: a Jira ticket, a document, a recording, or a Slack thread — or a blank plan with nothing inferred. Studio reads the source and proposes acceptance criteria as a table, each one marked:

• Stated — pulled verbatim from the source (an AC bullet, a line in a doc).
• Suggested — inferred by Studio, and never silent about why: every suggestion carries the reason it was raised (e.g. "61% of guest checkouts on staging come from a phone, and no existing plan covers mobile checkout").

Nothing is generated from a criterion until a person keeps it. An unconfirmed suggestion never reaches the plan.

2. The plan

The plan document itself: prose context (what's being tested and why, test data, where it runs, risks, what's explicitly out of scope, an activity feed) plus the acceptance criteria table — the part everything downstream keys off.

Each criterion carries:

Criteria can also be added by hand at any point — a plan is not required to start from a ticket.

3. Traceability

The full matrix: one row per test, grouped by the criterion it proves, running from the source line it came from to the sign-off it needs. The row that matters most is the one with nothing in it — a criterion with no test still gets a row, and a criterion whose surfaces aren't all covered gets a second one, rather than the gap being invisible because nothing rendered.

4. Build

For each criterion still missing coverage, Planner proposes a test and is explicit about what it will actually do:

Proposals live in the session only — the repo is untouched until you confirm which ones to use. Confirmed proposals can be handed to the coding agent as a build brief, or built by hand.

5. Run

A scope picker, not a new test runner — Planner hands the selected test ids to the same run path Tests already uses, and Infrastructure still owns the execution targets. Scope options:

• Everything
• P1 criteria only
• What changed — tests Studio itself linked since the last run
• What failed last time

Progress is counted in criteria decided, not files finished — the number a tester actually reports in stand-up.

6. Report

One line at the top answers the only question being asked — can we ship? — followed by the verdict per criterion, evidence counts, and whether the run's results actually reached your TestRelic account:

That last distinction is deliberate: Studio only claims an upload when it has a receipt for one. A run it can't confirm is reported as unconfirmed, not silently marked as failed.

7. Share

Two things to decide: who can see it, and what travels with it.

• Who — everyone at your organisation, anyone with the link (no sign-in — for a client or an auditor), or only you.
• What — the plan and its criteria, the verdict per criterion, and evidence (screenshots/video/traces) are on by default. Network and console logs, and test code, default off — sharing a result is not the same act as sharing your internals, so turning either on has to be deliberate.

Verdicts

A criterion's verdict for a given run is one of:

not_proven is the load-bearing value here: a criterion nothing ran against is not a pass and not a failure, and Planner never rolls it up as green. Sign-off is blocked by any unmet or unproven criterion, regardless of priority.