Umbra Group / Studio / Handoff Playbook
← Build v1.0 · Phase 4
DocumentHandoff Playbook
ClassificationInternal
Versionv1.0 · 2026.04
OwnerUmbra Studio
Phase 4 · Handoff · Two Weeks + 30-Day Support

The Handoff Playbook.

Day-by-day operational playbook for Weeks 10–11 of a Lighthouse Sprint and the 30-day support window that follows. Covers the training curriculum, supervised operation, outcome measurement, pattern harvest, and the Gate 4 independence verification — the phase where Studio walks out of the room and the client owns the system.

Paired with the Operations Manual, Governance Runbook, Risk Register, and Outcome Report templates. Read after the Build Playbook; keep open Days 46 through the end of the support window.

§01

How to use this playbook.

Audience · reading path · relationship to Build and the support window

Handoff is the transfer phase of a Lighthouse Sprint. Build produced a running system; Handoff produces an independently owned system. The work is not construction — it is training, confidence-building, outcome measurement, and pattern extraction.

The playbook is written for the Sprint Lead, who during Handoff is simultaneously the teacher and the auditor. The Sprint Lead owns the training curriculum, runs the supervised operation, authors the Outcome Report, and is the named signatory of the Gate 4 decision. The Agent Engineer and Governance Architect remain present but retreat further each day — by Day 55, neither should be answering a question the Workflow Owner cannot answer themselves.

If Gate 3 passed with Conditional Pass, every condition must be closed by Day 47. Training cannot compete for the team’s attention with unfinished Build work.

Inputs required on Day 46

Artifact from BuildUsed for
Running agent rosterThe system Handoff is transferring — must be stable from Day 46 onward.
Governance Wrapper v1.0Workflow Owner is trained on each of the six components; Technical Counterpart inherits configuration.
Gate 3 deck & decision memoBaseline for training scope; any Conditional Pass items drive the first three days of Week 10.
Risk Register v2.1Every open risk becomes a training topic; discharged risks become case studies.
Pattern Library candidatesPatterns confirmed or abandoned in the final pattern harvest (§06).
Drill reportThe Week-8 incident response transcript is training material for how to handle live incidents.
72-hour autonomy logReference — the Workflow Owner is about to own a system that already proved it can run itself.
§ Role inversion
In Build, the Studio team drove the system and the Workflow Owner watched. In Handoff, the Workflow Owner drives and the Studio team watches. If the Sprint Lead catches themselves reaching for the keyboard on Day 50, they have already failed the Gate 4 test.

Reading order

Read §02 and §03 before Day 46. §04 is read Sunday before Week 11 starts. §05 (Outcome Measurement) is the framework the Sprint Lead uses to author the Outcome Report across Week 11 — start reading it Day 48. §06 is the pattern harvest session, held Day 54. §07 is Gate 4. §08 covers the 30 days after Gate 4. §09 is Indietheka. §10 is the weekly checklist.

§02

Handoff at a glance.

Two-week arc · 30-day support tail · primary deliverables

Ten working days on site (Week 10 and Week 11), followed by a 30-day support window — email-only, four named office hours, capped Studio time. The centre of gravity shifts once per week: Week 10 is training, Week 11 is supervised operation, and the support window is availability without presence.

The two-week + 30-day arc

PhaseActivityWho drivesExit state
Week 10 · Days 46–50Training & transfer — Operations Manual walkthrough; state spreadsheet ownership transfer; governance drills rehearsedSprint Lead (teaching) · Workflow Owner (learning)Workflow Owner can operate the system unassisted on common paths
Week 11 · Days 51–55Supervised operation — Workflow Owner runs the system live, Studio team in the room but not touching controlsWorkflow Owner · Studio silentSystem runs five consecutive days under Workflow Owner authority without Studio intervention
Day 55 (Fri)Pattern harvest + Gate 4 — final pattern extraction session, Outcome Report presentation, Gate 4 decisionSprint Lead · full stakeholder groupGate 4 decision rendered; engagement closes; 30-day support window opens Monday
Days 56–8530-day support window — email-only; four 60-min office hours across the month; Studio answers, Workflow Owner ownsWorkflow Owner (ops) · Studio (advisory)Support window closes; engagement archived; post-engagement survey returned

Deliverables produced during Handoff

  1. Operations Manual v1.0 (lighthouse-operations-manual.docx) — the canonical document the Workflow Owner uses in perpetuity. Authored during Build, finalised Day 48.
  2. Training certificate — one-page record signed by Sprint Lead, Workflow Owner, and Executive Sponsor attesting the training curriculum was delivered and received. Signed Day 50.
  3. Outcome Report (lighthouse-outcome-report.docx) — the engagement’s headline document: baseline vs. post-Build metrics, patterns extracted, narrative of the sprint. Drafted across Week 11, presented Day 55.
  4. Confirmed Pattern Library entries — any candidates elevated to USP-### status, with full pattern cards for each. Written Day 54.
  5. Risk Register v3.0 — final register; any open-with-treatment items handed to the Workflow Owner with treatment plans. Closed Day 55.
  6. Gate 4 decision memo — one page, signed by Workflow Owner and Executive Sponsor. Signed Day 55.
  7. Post-engagement survey — completed by the client at Day 85, archived with the engagement. Distributed Day 85, target return Day 92.
§ Handoff is not training
Training is one of four activities in Handoff. The others are supervised operation, outcome measurement, and pattern extraction. If the Sprint Lead treats Handoff as a two-week training course, three of the four deliverables will be missing at Gate 4.
§03

Week 10 · training & transfer.

Days 46–50 · curriculum · transfer of authority

Five days to move every operational responsibility from Studio to the client. The structure mirrors an apprenticeship: the Sprint Lead demonstrates, the Workflow Owner mirrors, the Workflow Owner performs alone. By Day 50, the Workflow Owner runs a complete day-in-the-life cycle without Studio touching the system.

Day 46 · Mon Handoff kickoff. Training framed · curriculum walked

09:30 – Handoff kickoff meeting (90 min). Sprint Lead frames the two-week arc. Workflow Owner is introduced as the new system operator. Studio roles reframe: Sprint Lead = trainer, Agent Engineer = referent for edge cases, Governance Architect = referent for governance drills. Any Conditional Pass items from Gate 3 are walked and assigned for Tuesday closure.

11:00 – Operations Manual walkthrough (2 hrs). Sprint Lead and Workflow Owner read the Operations Manual together, cover to cover. Workflow Owner annotates: “I understand this,” “I need a demo of this,” “this section is unclear.” Manual is edited in real time.

14:00 – State spreadsheet tour (90 min). Workflow Owner sits at the controls. Sprint Lead narrates each column, each pill, each writeable vs. read-only boundary. Workflow Owner drives cursor and keyboard; Sprint Lead does not touch the screen.

Outputs:

  • Operations Manual edit list — items to clarify before Day 48 freeze.
  • List of live demos Workflow Owner requested — one per identified knowledge gap; scheduled for Days 47–48.
  • Conditional Pass closure plan — owner and due-date per item.
Day 47 · Tue Agent-by-agent walkthrough. Per-agent demonstration · failure drills

09:30 – Agent roster walkthrough (full day). One agent at a time. Sprint Lead demonstrates a live invocation; Workflow Owner mirrors; Workflow Owner runs the agent alone. For each agent the demo covers: normal operation, how to read the state-spreadsheet output, how to recognise bad output, how to pause the agent via Override, how to audit prior runs.

Time budget: 30–45 minutes per agent. Ten agents = 5–7 hours of training; allow whole day.

Agent Engineer present but silent. The Sprint Lead drives the curriculum. If the Workflow Owner asks a question the Sprint Lead cannot answer, the Agent Engineer answers — and the Sprint Lead adds the answer to the Operations Manual before end of day.

Outputs:

  • Completed agent walkthrough log — one signed line per agent (Workflow Owner + Sprint Lead).
  • Operations Manual updated with any missing how-to content.
  • A “Workflow Owner bench” — the list of agents the Workflow Owner is comfortable running alone vs. those requiring further rehearsal.
Day 48 · Wed Governance drills rehearsed. The six components exercised live

09:30 – Governance walkthrough (full morning). Governance Architect leads; Sprint Lead and Workflow Owner participate. Each of the six components is exercised as a live drill:

  • Monitoring — Workflow Owner opens the dashboard, identifies a degraded metric, narrates what it means.
  • Alerting — Governance Architect fires a test alert at each SEV level; Workflow Owner confirms the right human received it.
  • Audit Trail — Workflow Owner answers a forensic question (“Show me every invocation of agent X last Tuesday”) using only the audit log.
  • Escalation — Workflow Owner walks the escalation matrix from memory.
  • Override — Workflow Owner pauses and resumes a live agent via the state spreadsheet.
  • Rollback — Workflow Owner initiates a rollback to a Week-8 snapshot and restores; Governance Architect confirms integrity.

14:00 – Operations Manual freeze. All in-flight edits merged. Manual is locked at v1.0 and distributed to Workflow Owner, Executive Sponsor, Technical Counterpart.

Outputs:

  • Governance drill log — each of 6 components signed off.
  • Operations Manual v1.0 locked and distributed.
  • Risk Register v2.2 — any concerns surfaced by drills captured as items.
Day 49 · Thu Full-day rehearsal. Workflow Owner drives · Studio silent

09:30 – Full-day operational rehearsal. Workflow Owner runs the system as they will in Week 11 — Monday-morning startup, mid-day check-ins, afternoon reviews, end-of-day closure. Sprint Lead watches; never touches the keyboard. When the Workflow Owner stalls, Sprint Lead coaches verbally.

Any gap is fixable today. If the Workflow Owner cannot perform an operation without verbal coaching, that operation gets a final rehearsal slot Friday morning. If the gap is systemic (e.g. the Operations Manual is wrong), fix the Manual today and re-lock.

Studio intervenes only for: system failure that is not part of training (a real alert that would have fired regardless of the rehearsal); a data integrity risk that Workflow Owner has not noticed. Verbal coaching is not intervention — taking keyboard control is.

Outputs:

  • Rehearsal log — Sprint Lead records every coaching moment and whether it was resolved during the day.
  • A short list of Friday-morning re-rehearsal targets, if any.
Day 50 · Fri Transfer of authority. Training certificate signed · Week 11 opened

09:30 – Final re-rehearsals on any items flagged Thursday. Typically under 90 minutes.

11:00 – Training certificate signing. Sprint Lead, Workflow Owner, and Executive Sponsor sign the one-page training certificate. Certificate records: agents covered, drills rehearsed, Operations Manual v1.0 delivered, Workflow Owner’s confidence statement.

14:00 – Week 11 framing. Sprint Lead reads the Week 11 protocol aloud: Workflow Owner drives; Studio present but silent; every intervention logged. Protocol includes the escalation path — when does Workflow Owner ask for help, and how?

15:00 – Formal handoff moment. The shared channel receives a single message, posted by Sprint Lead, acknowledged by Workflow Owner: “The system is now under Workflow Owner authority. Monday 09:30.”

Outputs:

  • Training certificate (signed PDF).
  • Week 11 protocol memo (posted in shared channel).
  • Closure of all Day-46 Conditional Pass items; updated Risk Register.
§ Teaching vs. doing
A Sprint Lead who cannot resist taking the keyboard during Week 10 is signalling that the Operations Manual is not good enough. Fix the Manual. Don’t rescue the rehearsal.
§04

Week 11 · supervised operation.

Days 51–55 · Workflow Owner owns · Studio audits

Five days of the Workflow Owner running the system as the system’s owner. Studio is in the room (or on the call), but silent. Every intervention by Studio is logged; the goal is zero interventions by Thursday. Outcome measurement runs in parallel — every agent invocation feeds the Outcome Report.

Day 51 · Mon Supervised day one. Workflow Owner drives · first full day

09:00 – Workflow Owner begins the day as Operations Manual prescribes. Sprint Lead observes; Agent Engineer and Governance Architect are on call but not present. End-of-day debrief at 17:00 (30 min) — Workflow Owner describes the day; Sprint Lead records observations.

Outcome measurement begins. Sprint Lead opens the Outcome Report draft; first throughput, quality, and error-rate data points from Week 11 start accruing.

If Studio intervenes: log the event in the Intervention Log with timestamp, reason, resolution, and the classification (critical system rescue / training gap / non-urgent assist). Non-urgent assists are disallowed — they count as failures of the Handoff.

Day 52 · Tue Supervised day two. Retreat begins

Studio retreats one step. Sprint Lead attends morning startup (30 min) and end-of-day debrief (30 min); otherwise, not in the room. Workflow Owner runs the full day alone.

Agent Engineer and Governance Architect are reachable by chat but not pre-deployed. Questions must be asked; no unsolicited support.

Sprint Lead’s day is spent authoring the Outcome Report draft based on yesterday’s data plus the running totals from Week 11 to date.

Day 53 · Wed Independent day. Studio unavailable except for true emergencies

Scheduled Studio-off day. Sprint Lead is available only by pre-declared office hour (60 min window, typically 14:00–15:00). Agent Engineer and Governance Architect are off the engagement entirely. Workflow Owner operates the system exactly as they will after Day 55.

This is the day the Gate 4 question is most visible: can the Workflow Owner, without Studio in the room, run a full day and produce the expected outcome volume?

Any Wednesday intervention is a serious signal. If the system cannot run for one day without Studio, it will not survive the 30-day support window in its current state.

Day 54 · Thu Pattern harvest. Final extraction session · Outcome Report polish

09:30 – Pattern harvest session (3 hrs). Full Studio team plus Workflow Owner. Walk every pattern candidate raised during Discovery, Redesign, and Build. For each: Confirm (assign USP-###), Abandon (with note), or Carry-forward (for a later sprint to confirm). See §06 for the structured template.

14:00 – Outcome Report polish. Sprint Lead presents the report draft to the full Studio team for internal critique before Friday. Workflow Owner receives a pre-read copy for overnight review.

16:00 – Week 11 intervention log review. Sprint Lead reads the intervention log aloud; classifies each as training gap (resolved or not) or genuine system issue. Classification feeds the Gate 4 decision.

Day 55 · Fri Gate 4 · independence verified. Outcome Report presentation · decision · closure

11:00 – Outcome Report presentation (90 min). Full stakeholder group: Workflow Owner, Executive Sponsor, Technical Counterpart, full Studio team. Sprint Lead presents the report; the room discusses; Workflow Owner renders the ready statement.

14:00 – Gate 4 decision (30 min). Per §07. One of: Pass, Conditional Pass, Hold. Decision signed by Workflow Owner and Executive Sponsor.

15:00 – Closing ritual. Sprint Lead hands the Workflow Owner the full engagement archive: Operations Manual, Governance Runbook, Outcome Report, Risk Register, confirmed patterns, training certificate, Gate 4 memo, drill report. Short verbal acknowledgement of the transition. Support window opens Monday (Day 56).

§ Silence is a deliverable
A Week 11 with zero Studio interventions is the single clearest signal that the sprint succeeded. If the intervention log is empty on Thursday, Gate 4 is a formality.
§05

Outcome measurement.

How to compare post-Build performance to the Discovery baseline

The Outcome Report is the engagement’s headline document. It is also the easiest document to write badly. Numbers are seductive; if the Sprint Lead is sloppy with baselines, sample sizes, or attribution, the report is anecdote dressed as evidence. This section is the discipline that prevents that.

The four outcome axes

Axis 01 ▸ THR

Throughput.

How many “units of work” per week? Defined in Discovery; measured the same way post-Build. Units per week, sustained.

Axis 02 ▸ QLT

Quality.

What proportion of outputs meet the Workflow Owner’s acceptance bar? Uses the same rubric as the Week-6 baseline comparison. Percent acceptable.

Axis 03 ▸ HHR

Human hours.

Hours of human work required per unit. Stopwatch protocol (D-02) re-run post-Build. Hours per unit.

Axis 04 ▸ ERR

Error rate.

Proportion of unit outputs that required rework post-release. Uses the Discovery definition of “error.” Percent reworked.

Measurement discipline

RuleWhat it enforces
Identical definitionsIf Discovery defined “throughput” as reviews-per-week, Outcome uses reviews-per-week — not pieces-per-week, not drafts-per-week. Definition drift invalidates the comparison.
Adequate sample sizesDiscovery baselines must have had n ≥ 20; Outcome measurements must meet or exceed that. Smaller samples are reported as directional only.
No cherry-picked windowsOutcome uses Week 11 data in its entirety — not “best week of Build.” If Week 11 was interrupted by a client holiday, extend into the support window and note the adjustment.
Attribution disciplineIf an outcome gain was caused partly by a business-side change (new hire, new tool, unrelated process fix), say so. Unattributed gains undermine the sprint’s credibility for future clients.
Confidence intervals where possibleIf sample sizes support it, report ranges not just point estimates. For n < 50, state n explicitly; for n ≥ 50 and normally-distributed metrics, attach a 95% CI.
Discovery-era numbers citedThe Outcome Report links directly to the Discovery Report as its baseline authority. Every number in the comparison is traceable.

Outcome Report · required structure

  1. Executive summary (1 page) — one-sentence headline per axis; overall recommendation.
  2. Context — client, engagement window, Discovery findings in one paragraph.
  3. Baseline — the Discovery numbers, re-stated with sample sizes and methods.
  4. Post-Build measurement — the Week 11 numbers with the same disclosure level.
  5. Delta — axis-by-axis comparison, with attribution notes.
  6. Narrative — 4–6 paragraphs walking the engagement’s story: bottleneck, design choice, key Build decisions, incidents navigated.
  7. Patterns extracted — table of confirmed USP-### entries with one-line descriptions.
  8. Risk Register v3.0 — open items carried forward, with treatment plans assigned to the Workflow Owner.
  9. Support plan — how the 30-day window will be used; office hour schedule.
  10. Recommendation & Gate 4 ask — Pass / Conditional Pass / Hold proposal.
§ Outcome != result
A sprint that hit its throughput target but saw quality fall is not a successful outcome — it is a tradeoff. Report all four axes, even when one is unflattering. Selective reporting destroys the report’s evidentiary value.
§06

Pattern extraction.

Day 54 · the harvest session · turning findings into USP entries

The Pattern Library is the strategic moat of Umbra Studio. Every sprint that passes Gate 4 contributes confirmed patterns; the library compounds across engagements. Pattern extraction is the single activity that makes every future sprint cheaper and better than the last.

A pattern is not a best practice. It is a named, reusable configuration (of agents, governance, and human-in-the-loop) that solved a repeatable problem, observed in at least two engagements, with a documented trigger, shape, and exception.

Confirmation bar — a pattern is confirmed when all four hold

CriterionDefined as
NameableA distinct noun phrase (“Spreadsheet-as-State-Machine”) not a verb (“communicate with stakeholders”).
Demonstrably effectiveThe pattern, as applied in this sprint, moved an outcome axis. “We used a spreadsheet” is not a pattern; “USP-009 allowed humans and agents to share authority over the same state without a separate admin UI” is.
Triggered reliablyA future Sprint Lead can identify when to reach for this pattern. Examples: “use USP-009 when the Workflow Owner must edit agent state directly.”
Observed elsewhereAt least one prior engagement or public analogue exhibits the same shape. First observation is a candidate; second observation is a pattern.

Pattern card · canonical format

USP-### · <Title>
One-line premise.

Trigger: When the Sprint Lead should reach for this pattern — 1–2 sentences.

Shape: What the pattern is — 2–4 sentences describing the agents, governance, and human-in-the-loop configuration.

Exception: When not to use this pattern — 1 sentence.

Observed in: List of engagements where the pattern was confirmed (earliest first).

Related patterns: USP-### cross-references where applicable.

The Day 54 harvest session · agenda

BlockDurationContent
Candidate inventory15 minAgent Engineer reads the list of pattern candidates raised during Discovery, Redesign, and Build. No discussion yet.
Per-candidate assessment90 minWalk each candidate against the four-criterion bar. Decide: Confirm (author USP card), Abandon (record why), Carry-forward (note for future sprint).
USP card authoring45 minFor each Confirm, Sprint Lead drafts the card in real time; Agent Engineer and Governance Architect critique until the card meets the canonical format.
Cross-reference20 minReview new USP entries against the existing library; add related-pattern links; update any older cards with new observations.
Close10 minFinal tally; Sprint Lead posts the list of new USP-### entries to the Studio archive and the Outcome Report appendix.
§ Scarcity discipline
A typical sprint produces 1–3 confirmed patterns. If the harvest is producing five or more, the bar is too loose; re-read the criteria. A tight library is more valuable than a long one.
§07

Gate 4 · independence verified.

Day 55 · the final decision · closure of the engagement

Gate 4 decides whether the engagement closes or extends. It is attended by Workflow Owner and Executive Sponsor on the client side; Sprint Lead and (for the formal sign) Studio principal attend on the Studio side. Outcome: Pass, Conditional Pass (support window starts with specific conditions named), or Hold (engagement extends by 1–2 weeks at no cost to the client, Handoff re-run).

Gate 4 · Independence Verified

Does the client own the system, with the knowledge, documentation, and confidence to run it unassisted under the 30-day support window’s terms?

The sprint passes Gate 4 when the system satisfies all of the following:

  • Training certificate signed. Delivered Day 50. No gaps flagged unresolved.
  • Week 11 intervention log clean. Zero non-emergency Studio interventions; emergency interventions classified and their root causes discharged. Weekly totals across the week must show a flat or declining trend, not a rising one.
  • Outcome Report accepted. Workflow Owner and Executive Sponsor agree the report is accurate and the numbers representative.
  • Operations Manual in use. The Workflow Owner has demonstrably used the Manual as their reference during Week 11 — not Studio staff.
  • Governance components fully exercised. All six drilled successfully in Week 10; at least four fired live or rehearsed in Week 11.
  • Risk Register v3.0 closed or accepted. Every risk has status — no open-silent items. The Workflow Owner owns treatment of any carried-forward items.
  • Pattern harvest completed. Day 54 session produced confirmed USP entries or a reasoned zero count.
  • Workflow Owner ready statement. Workflow Owner answers on the record: “I own the system. I can run it for thirty days with only email support.” A “no” triggers Hold (not Conditional Pass — Gate 4 Conditional is narrow and formal).

Decision outcomes

Outcome 01 ▸ PASS

Engagement closes.

Support window opens Monday. Outcome Report published (internally). Payment milestone M5 released. Client is a reference candidate pending 30-day outcome hold.

Outcome 02 ▸ CONDITIONAL

Close with named items.

Engagement closes but support window starts with specific, dated items assigned to Studio for email resolution — typically documentation gaps or single-agent tunings. Used sparingly.

Outcome 03 ▸ HOLD

Extend by one week.

Engagement extends by up to two weeks at no additional cost to the client; Handoff re-runs the failed component. If Hold recurs, a post-mortem is commissioned and the engagement scope is renegotiated.

§ Payment closure
Gate 4 Pass or Conditional Pass releases payment milestone M5 (15% of engagement fee). Hold defers M5 until Gate 4 is re-run successfully. M5 is the only milestone tied to Gate 4 to keep both parties aligned on closure discipline.
§08

The 30-day support window.

Days 56–85 · email-only · capped · deliberate

The 30-day support window is deliberately narrow. It is not a retainer. It is not ongoing ops. It is the period during which the client is settling into the system and Studio is available, at a limited cadence, to answer questions and patch documentation gaps. The narrow shape protects both parties.

The four terms

TermDefinition
Email-onlyAll support requests go to a single Studio address. No Slack, no ad hoc calls. Studio responds within two business days.
Four office hoursFour 60-minute video calls scheduled across the 30 days (Days 60, 67, 74, 81 typical). Workflow Owner brings questions; Sprint Lead answers.
No code changesSupport is advisory. Studio does not modify agents or governance during the support window — if a change is needed, it is scoped as a new engagement. Exceptions require Studio-principal sign-off.
Capped at 12 hoursTotal Studio time in the window is capped at 12 hours (4 office hours + 8 hours of email-and-triage). Overflow is either out-of-scope or converted to a new engagement.

Support-window calendar · typical

DayEventPurpose
Day 56 · MonSupport window opensSprint Lead posts the four office-hour slots and support address in the shared channel.
Day 60 (≈)Office hour 1First week shakeout: anything the Workflow Owner encountered in the first five days of independent operation.
Day 67 (≈)Office hour 2Risk Register review: any treatment items the Workflow Owner needs help completing.
Day 74 (≈)Office hour 3Pattern observations: new patterns the Workflow Owner is seeing in operation; candidate for future sprint or library update.
Day 81 (≈)Office hour 4Pre-close: survey distributed; any final documentation gaps closed; handoff of support archive.
Day 85 · FriSupport window closesSupport address auto-replies with a “support window closed” message and redirects to the engagement archive.
Day 92 (target)Post-engagement survey returnedWorkflow Owner and Executive Sponsor return a short survey; Sprint Lead archives with the engagement record.

What support is and is not

Support ▸ IS

Answering questions.

How do I interpret this alert? What does this audit row mean? Where in the Manual does this case live? Reference help, not hands-on work.

Documentation clarifications. Pattern questions. Post-mortems of minor incidents. Confirming when a situation requires a new engagement vs. self-service.

Support ▸ IS NOT

Running the system.

Studio does not execute agents, edit the state spreadsheet, triage alerts, or modify code. These are the Workflow Owner’s job — always.

Adding net-new capability. Running agents that fall outside the Gate 2 contract. Any on-call responsibility. All of these are new-engagement territory.

§ Why the cap matters
Uncapped support becomes a zombie retainer. The 12-hour cap keeps the engagement closed, which is good for the client (they own what they paid for) and good for Studio (we can deploy the Sprint Lead to the next engagement on schedule).
§09

Worked example · Indietheka.

Two weeks of Handoff — plus 30 days in the rearview

Indietheka’s Handoff ran Days 46–55 of the internal sprint, with the support window running Days 56–85. The Workflow Owner (Abe) was the sole operator; Handoff therefore doubled as a sole-operator stress test of the Operations Manual. Gate 4 passed on Day 55 with no conditions attached. The 30-day support window closed quietly — four office hours used, 9.5 of 12 support hours consumed.

Week 10
Training

Day 46 opened with a 90-minute framing and a 2-hour Operations Manual read-through. The Manual was edited 18 times in real time; by Day 48, v1.0 was locked at 47 pages including appendices. Agent-by-agent walkthroughs on Day 47 took 5 hours 40 minutes — 34 minutes per agent average, with the Editorial Writer taking the longest (58 min) due to Spanish-language nuance coaching.

Governance drills on Day 48 exercised all six components cleanly. Rollback drill restored a Week-8 snapshot in 7 minutes 12 seconds. Day 49 rehearsal had three verbal coaching moments — all resolved without Studio touching the keyboard. Training certificate signed Day 50 at 11:24.

Week 11
Supervised

Day 51 ran clean: zero Studio interventions; five review drafts shipped through the agent roster. Day 52 saw one verbal coaching request (audit log forensic question); resolved by referring the Workflow Owner to page 31 of the Manual. Day 53 — the independent day — ran without any coaching contact; the pre-declared 14:00–15:00 office hour went unused.

Day 54 pattern harvest: 6 candidates walked. Four confirmed as USP entries (USP-009 Spreadsheet-as-State-Machine, USP-011 Weekly Demo, USP-013 72-Hour Autonomy Window, USP-014 Pre-Condition Guards). Two abandoned (one as “merely a best practice, not a pattern”; one as “observed only once, needs a second engagement”). The four confirmed patterns became the seed of the published Umbra Studio pattern library.

Day 55
Gate 4

Outcome Report headline: 7.2× throughput (6 reviews-per-week sustained vs. 0.8 baseline); 92% human-hour reduction (0.5 hrs/review vs. 6.4); error rate from 22% to 4%; cycle time from 5.2 days to 7 hours. Week 11 intervention log: zero non-emergency interventions, two verbal coaching moments, one independent day with no contact. Workflow Owner ready statement given verbatim, on the record.

Gate 4: Pass. Engagement closed Day 55 at 15:47. Closing archive delivered. Payment milestone M5 released.

Days 56–85
Support window

Four office hours used: Day 60 (weekly shakeout, 58 min); Day 67 (risk review, 47 min); Day 74 (pattern observations — Workflow Owner proposed USP-015, carried forward to future sprint for confirmation); Day 81 (pre-close, 32 min). Email support: 11 tickets, 9.5 total hours consumed. Zero code changes made during the window.

Post-engagement survey returned Day 89. Net promoter score 9/10. The Workflow Owner’s verbatim: “Handoff was the part of the sprint where I could feel the difference — I owned it and I knew I owned it.”

§ Self-audit
In the Indietheka engagement, Studio and client were the same person. This made Handoff unusually legible — every Studio-hat-off moment was visible. The lessons ported directly into this playbook: the Operations Manual freeze-date, the intervention log, the independent day, the capped support window.
§10

Handoff Lead checklist.

Run this every Monday of Handoff

Ten items for each of Week 10 and Week 11, plus a five-item support-window checklist.

Week 10 · Monday

  1. Gate 3 Conditional Pass items (if any) have named owners and Tuesday closure dates.
  2. Operations Manual is present, current, and distributable — even if edits are expected during the week.
  3. Agent walkthrough schedule for Day 47 is drafted with time-boxed slots per agent.
  4. Governance drill scenarios for Day 48 are defined; Governance Architect is ready to lead.
  5. Training certificate template is ready for Day 50 signing.
  6. Workflow Owner has full-day availability for Days 46–49 confirmed.
  7. Agent Engineer and Governance Architect know their silent-observer role and the coaching-vs-keyboard-taking boundary.
  8. Week 11 protocol memo is pre-drafted, ready to post Friday.
  9. The state spreadsheet schema is locked — no changes from here to Gate 4 without Studio-principal sign-off.
  10. The Outcome Report draft shell is open in the Studio archive folder.

Week 11 · Monday

  1. Intervention Log is open and blank for Monday entries.
  2. Office hour for Day 53 is pre-declared in the calendar (14:00–15:00 typical); Workflow Owner has the joining link.
  3. Outcome measurement data collection is wired — Week 11 metrics flow into the Outcome Report automatically.
  4. Pattern candidate list is consolidated ahead of Thursday’s harvest session.
  5. Pre-read copy of the Outcome Report is scheduled for Thursday delivery to the Workflow Owner.
  6. Day 55 Gate 4 meeting is on the Executive Sponsor’s calendar (11:00 presentation + 14:00 decision).
  7. Closing archive folder is drafted — empty but structured for the Day 55 handover.
  8. 30-day support window start-message is drafted for Day 56 Monday posting.
  9. Office hour slots for Days 60, 67, 74, 81 are held provisionally on the Sprint Lead’s calendar.
  10. M5 invoice draft is prepared, ready to send upon Gate 4 Pass.

Support window · ongoing

  1. Support email address is monitored daily (response SLA: 2 business days).
  2. Office hours go un-rescheduled except for emergencies — rhythm matters more than convenience.
  3. Every support ticket is classified: documentation, pattern, advisory, or new-engagement-candidate.
  4. Support-hour burn-down is tracked against the 12-hour cap; the Workflow Owner is notified at 8 hours, 10 hours, and cap.
  5. Day 81 office hour includes the post-engagement survey distribution and a verbal reminder of support window close.
§ Closing principle
The best Handoff is the one you don’t remember. Quiet weeks, clean certificates, honest numbers, a Workflow Owner who feels the system is theirs. If that happens, the Pattern Library gets richer, the client becomes a reference, and the next sprint starts with a stronger foundation than this one did.