Handoff: what the fictional client receives (fixture run, U24-B)
This is a fictional test artifact. The client does not exist; nothing here is a Backbrief decision. Written by the orchestrator on 2026-09-10 at about 05:05 EDT, after the verifier record.
What was done
One request in the brief (../fixture/fixture-brief.md, lines 9 to 12: referrals or a self-serve product for the next 90 days) went through the demonstrated workflow inside this folder:
/intakestructured the brief (01-brief-from-intake.md); every figure traces to the raw brief and the fields the brief does not state are marked not stated or unknown rather than filled.- The orchestrator wrote the scope and one plan unit before any work ran (
02-03-plan-unit.md, units F1 to F4 with owners, dependencies and pass or fail criteria). - The builder produced the deliverable: a one-page decision memo comparing more accountant referrals, the $79 a month self-serve product, and no change, with every number cited to a brief line and evidence tagged (
04-decision-memo.md, 676 words, then three wording corrections). - A verifier that had not seen the builder's reasoning checked the memo against the unit's criteria (
05-verifier-record.md): PASS, one confirmed wording defect (applied), four plausible notes (two applied as cite corrections, two left for the owner).
What the client receives
04-decision-memo.md, the memo, with its recommendation (extend the referral channel, hold the $500 in reserve) and the GO or NO-GO question the owner answers.05-verifier-record.md, the separate check that goes with it, so the client sees what was verified and what was not.
Open items and next action
- Two judgment calls the verifier flagged and the orchestrator left in place: the memo describes four of six cleanups from one accountant as a rate the owner might sustain (tagged assumption), and it treats the $500 ad budget as belonging to the product option when the brief states the budget on its own. The owner reads both before deciding.
- Five questions
/intakewould have asked are listed at the end of01-brief-from-intake.md; none was answered, because the input is a fixture. - Next action: the owner's GO or NO-GO on the memo's question. If GO on Option A, the next unit would be the referral plan itself, which this run did not produce by design.
Where the six artifacts are
| # | Artifact | File |
|---|---|---|
| 1 | The fictional brief | ../fixture/fixture-brief.md (raw) and 01-brief-from-intake.md (structured) |
| 2 | The scope | 02-03-plan-unit.md, "Scope" |
| 3 | The plan unit | 02-03-plan-unit.md, the table |
| 4 | The produced artifact | 04-decision-memo.md |
| 5 | The verifier record | 05-verifier-record.md |
| 6 | The handoff | this file |
How this run differs from a buyer's first run: the same installed agents ran in the developer's own session, by orchestrator dispatch, not from a clean-machine Claude Code session following ../FIRST-RUN.md. A buyer's run goes through /intake and the orchestrator as commands; the artifacts it should produce are the same six, judged by ../fixture/expected-artifacts.md.