D3 Sprint Operating Runbook v1.0
A sprint at D3 — who does what, when
Day 0 of 3
Sprint cycle
Sprint #42
Day 1 of 10
+3 commits
Feature A 50%
+5 commits · CI green
Feature A merged
Tests +12 · cov 82%
Feature B in QA
+4 commits
Demo prep
Refinement
Week 0
spec.md + AC
Planning
Day 1
Sprint goal committed
Demo
Day 9
Demo recording + accept log
Retro
Day 9 PM
Action items + lessons
Ticket lifecycle inside a sprint · 3-day SLA
Day 0
Pre-sprint
Day 1
Plan
Day 2
Build
Day 2.5
Test
Day 3
Deploy
PO
Triage
NEW -> Backlog
Sprint goal
Commit scope
Accept
Per story in ADO
accept log
BA
Refine
Spec + AC + edges
spec.md + AC
Demo prep
Script + agenda
Close
Update KH
KH updated
Tech Lead
Risk flag
Sizing + DoR
Tech plan
Stack + split
Approve
Risky merges
Dev
Pick + branch
Start by lunch
Build + PR
spec: prefix · #ticket
PR opened
QA fixes
Address feedback
Deploy
Tag release · prod
release tag
QA
Test plan
Risk + scenarios
Cases curated
Manual + auto setup
Gatekeeper
Sign-off stage
qa-report.md
Smoke
Production check
AI Pipeline
E2E pipeline
5-stage parallel
Allure
Auto signal
Block / proceed
PO
BA
Tech Lead
Dev
QA
AI Pipeline
output artifact
1. Sprint at a glance
Hard SLA — 3 working days from TO DO → DONE
Pre-sprint refinement is rolling (Week 0) and does not count toward the 3-day budget. The clock starts when a ticket leaves READY for TO DO and stops when it lands in DONE. Overdue tickets trigger a root-cause item in the next retro.
Week 0 · rolling
Pre-Sprint Refinement
- Intake of new requests
- Spec drafting (SpecKit + Claude Code)
- Sizing + edge cases
- DoR sign-off — all stories READY by Day 0 PM
Does NOT count toward 3-day SLA
Day 1 · 3.0 d left
Sprint Planning
- Sprint goal committed
- ADO subtasks created
- Scope locked
- Pick + start by lunch
Hours 0–8
Day 2 · 1.5 d left
Development
- Branch + TDD + AI unit tests
- PR opened AM —
spec:prefix +#ticketfooter - Code review ≤ 4 hours
- Merge by PM
Hours 8–16
Day 2.5 · 0.5 d left
Continuous Testing
- 5-stage E2E pipeline (parallel)
- QA manual + AI automation in parallel
- QualityGatekeeper sign-off — ONLY HUMAN gate
- Stage release ready
Hours 16–20
Day 3 · HARD STOP
PR & Release
- Demo AM
- PO accepts per story
- Deploy to production
- Smoke check + close ticket
Hours 20–24 · clock stops in DONE
AI does the heavy lifting
SpecKit drafts specs, Copilot writes unit tests, Playwright+AI runs E2E, automated pipeline triggers on every PR. Humans focus on judgment.
One human gate, not many
QualityGatekeeper is the only mandatory human checkpoint. Everything else flows on automated signal. This keeps the 3-day SLA realistic.
DoR is sacred
If a story isn't READY by Day 0 PM, it does not enter the sprint. Half-baked refinement is the #1 cause of SLA overrun.
2. Roles in the sprint
Six roles run a sprint. The min level shown is the floor — anyone above it qualifies. People below the floor shadow but do not own.
PM / Scrum Master
Min PM2 to chair sprint
Owns the sprint cadence end-to-end. Chairs planning + demo, facilitates standup + retro, monitors SLA budget, escalates blockers. PM1 shadows; PM3+ for multi-team sprints.
Product Owner (PO)
Client-side or BA3 proxy
Owns the backlog ranking and acceptance. Usually client-side. If client cannot dedicate a PO, a BA3 (Client-Facing track) can proxy with documented client trust.
Business Analyst (BA)
Min BA2 to chair refinement
Chairs refinement, drafts spec + AC + edge cases via SpecKit, owns Knowledge House updates, runs client-facing meetings within the sprint. BA1 contributes content but does not chair.
Tech Lead
Min SE L4 (Tech Lead track)
Wears two hats: technical decision-maker for the sprint AND active engineer when capacity allows. Curates Custom Skills + Constitution, gates risky merges, owns architectural calls. Solo gatekeeper for any L5-level (SA / Distinguished Engineer) decisions.
Developers
SE L1 ships paired; L2+ ships solo
Implement features end-to-end inside the sprint. L1 ships with paired review by L2+. L2 owns small features solo. L3+ gates code review. L4 Tech Lead approves architectural changes.
QualityGatekeeper (QA)
Min QA3 to gate; QA2 covers manual
The only mandatory human gate in continuous testing. QA3 (Analyst or SDET track) signs off stage releases. QA2 owns manual exploratory + UX/risk testing. QA1 paired with QA2+.
3. Ticket lifecycle — 12 states
Click any state to expand. Each card shows the owner, the minimum level to own it, entry/exit criteria, SLA budget impact, and the artifacts produced.
1
NEW
Owner: PO
Min level to own
PO (client or BA3 proxy)
Entry criteria
- Request created in ADO with title + initial description
Exit criteria
- PO has triaged + assigned to BACKLOG within 48h
SLA budget
Not yet started · 3.0d intact
2
BACKLOG
Owner: PO + PM
Min level to own
PO ranks · PM2+ assists with commercial framing
Entry criteria
- PO has triaged the request and ranked it by business value
Exit criteria
- Selected for the next refinement session (by capacity)
SLA budget
Rolling · clock not started
3
IN REFINEMENT
Owner: BA (chair)
Min level to own
Entry criteria
- Story selected from BACKLOG with PO context attached
Exit criteria
- Clarified spec drafted in SpecKit (preplan.md + spec.md)
- Acceptance criteria agreed by BA + Dev + QA
- Edge cases listed
- Initial sizing done
SLA budget
Pre-sprint · clock not started
4
READY
Owner: BA + Tech Lead
Min level to own
Entry criteria
- Refinement outputs complete (spec + AC + edges + sizing)
Exit criteria — Definition of Ready
- Story meets DoR — sized, AC clear, FR-traceability matrix updated
- Test data + environments confirmed ready for AI pipeline
- Story enters sprint on Day 0 PM at latest
SLA budget
Pre-sprint · 3.0d intact
5
TO DO
Owner: PM (assigns) + Dev (claims)
Min level to own
Entry criteria
- Sprint planning completed · scope locked · ticket assigned
Exit criteria
- Developer has branch created + first commit pushed
SLA budget
⏱ Clock STARTS · 3.0d remaining
6
IN PROGRESS
Owner: Dev
Min level to own
Entry criteria
- Branch + first commit · subtasks expanded in ADO
Exit criteria
- TDD: unit tests written + passing locally
- AI unit tests generated via Copilot + self-tested
- PR opened with
spec:prefix +#ticketfooter
SLA budget
End of Day 2 AM · ~1.5d remaining
7
IN CODE REVIEW
Owner: Reviewer (peer + Tech Lead + AI)
Min level to own
Entry criteria
- PR opened with correct convention + Copilot review attached
Exit criteria
- Peer review approved (≤ 4 hours target)
- AI review issues resolved or explicitly accepted
- Merged to main · Done-dev gate ticked in ADO
SLA budget
End of Day 2 PM · ~1.0d remaining
8
IN QA (Manual)
Owner: QA — exploratory + risk + UX
Min level to own
Entry criteria
- Merge complete · staging deploy successful · test cases curated
Exit criteria
- Exploratory pass complete · risk + UX issues filed
- Pass: screenshots attached to user story · Fail: bug linked to story (dedup)
SLA budget
Day 2.5 · 0.5d remaining
8b
IN AUTOMATION (parallel)
Owner: AI Pipeline + QA SDET
Entry criteria
- PR triggers 5-stage E2E pipeline · Playwright + Allure
Exit criteria
- All 5 stages pass
- Allure report + qa-report.md attached to PR
- QualityGatekeeper (QA L3) signs off
SLA budget
Day 2.5 · 0.5d remaining
9
READY FOR DEMO
Owner: BA + Tech Lead + QualityGatekeeper
Min level to own
Entry criteria
- All manual + automated tests pass on staging
Exit criteria
- Stage release tagged · demo agenda confirmed
- Test data + access provisioned for PO
SLA budget
Day 2.5 EOD · 0.5d remaining
10
ACCEPTED
Owner: PO + PM
Min level to own
Entry criteria
- Demo delivered · demo recording attached · ADO feedback log captured
Exit criteria
- PO marks Accepted in ADO per story (block/proceed decision)
- Any rejected story re-enters BACKLOG with PO feedback
SLA budget
Day 3 AM · ~3 hours remaining
11
DEPLOYED
Owner: Dev / Tech Lead
Min level to own
Entry criteria
- PO acceptance + ADO pipeline triggered + release notes generated
Exit criteria
- Production deploy complete · release tagged
- Smoke check passes within 1 hour of deploy
SLA budget
Day 3 EOD · < 2 hours remaining
12
DONE
Owner: BA closes · PM verifies
Min level to own
Entry criteria
- Smoke check passed in production
Exit criteria
- Ticket closed · Knowledge House updated · release notes published
- If overdue: root-cause item added to next retro
SLA budget
🏁 Clock stops · 0d remaining
4. Sprint ceremonies
Five ceremonies anchor the rhythm. Each one shows who chairs (with min level), who attends, time-box, and expected outputs.
Refinement (rolling, Week 0)
| Chair | BA2+ — drafts spec via SpecKit |
|---|---|
| Attendees | BA + Dev (L2+) + QA (L2+) + Tech Lead (consult) |
| Time-box | 60–90 min per session · 1–2 sessions per week |
| Outputs | preplan.md · spec.md · AC · edge-case list · sizing · FR-traceability matrix |
| DoR gate | All stories READY by Day 0 PM — non-negotiable |
Sprint Planning (Day 1 AM)
| Chair | PM2+ — facilitates capacity + scope commitment |
|---|---|
| Attendees | Full team + PO |
| Time-box | 2 hours max · pick + start by lunch |
| Outputs | Sprint goal · ADO subtasks · assignments · scope locked |
| Anti-pattern | Re-opening refinement during planning · auto-disqualifies story for this sprint |
Daily Standup
| Chair | PM (any level) facilitates · PM1 acceptable for routine standups |
|---|---|
| Attendees | Whole team · PO optional |
| Time-box | 15 min sharp · blockers + SLA-at-risk only |
| Outputs | Blockers logged in ADO · escalations triggered |
| Anti-pattern | Status round-the-room ("what I did yesterday") · use ADO board for that |
Demo (Day 3 AM)
| Chair | PM2+ runs format · BA2+ presents AC trace |
|---|---|
| Attendees | Full team + PO + invited stakeholders |
| Time-box | 60 min · 5–10 min per story |
| Outputs | Demo recording attached to ADO · feedback log · PO acceptance decision per story |
| Anti-pattern | Demo of half-built work · skipped if any story is below READY-FOR-DEMO |
Sprint Retro (after close)
| Chair | PM3+ facilitates · or external facilitator if PM is the subject |
|---|---|
| Attendees | Whole team · PO optional |
| Time-box | 60 min · action-item driven |
| Outputs | Action items in ADO with owners · root-cause items for every SLA overrun · trend signal flagged if same item appears 3 retros in a row |
| Anti-pattern | Round-the-room venting · no action items · "we should communicate more" written down again |
5. Quality gates
Six gates govern what flows. If you are below the listed min level, you can co-pilot but cannot be the gate.
1
Definition of Ready (DoR)
Story leaves refinement and enters sprint only if spec + AC + edges + sizing are all signed off. No exceptions.
2
PR convention compliance
Every PR must use the
spec: prefix + #ticket footer. PRs without convention get auto-rejected — Tech Lead does not waste cycles on form.3
Code review gate (Done-dev)
Merge to main requires peer L3+ approval + resolved Copilot issues. Tech Lead joins for any architectural surface — schema, public API, infra.
4
QualityGatekeeper — the only human gate
In Continuous Testing, automation runs everything but the final block/proceed decision is one human's signature. QA L3 reviews the pipeline output (Allure + qa-report.md) and stamps stage release. If QA L3 unavailable: backup is the Tech Lead, but they must flag it as a known gap.
5
Demo acceptance
PO marks block/proceed in ADO per story. PO rejection sends the story back to BACKLOG with documented feedback — it does not auto-fail the sprint.
6
Production smoke check
Within 1 hour of production deploy, the deploying engineer must run the smoke check and confirm in ADO. If smoke fails: rollback runbook (covered in v1.1).
6. Toolchain
The tools that make 3-day SLA realistic. AI does the heavy lifting; humans focus on judgment.
SpecKit + Claude Code
BA · Tech Lead · Dev
Refinement spec drafting · implementation pre-plan · Dev coding assist
MANDATORY
Azure DevOps + ADO MCP
Everyone
Ticket source of truth · sprint board · pipelines · release notes
MANDATORY
Figma + Figma MCP
BA (reads) · Dev (consumes)
UI/UX specs attached to user story
When UI story
GitNexus
Dev (primary)
Branching · PR · merge · release tagging
MANDATORY
Context7
SE L2+
Code understanding · cross-file reasoning during dev or review
Encouraged
Custom Skills + Constitution
Tech Lead curates · Dev consumes
Guard rails for AI coding · architectural patterns enforced
MANDATORY
Playwright + Allure
QA SDET (owns) · Pipeline (runs)
5-stage E2E pipeline · triggered on every PR
MANDATORY
7. Level × Sprint responsibility map
The single answer to "I am role X at level Y — what can I own in a sprint?" Columns = roles. Rows = ticket states. Each cell shows the minimum level to own that step plus the activity.
| State | PM | BA | Dev (SE) | QA | PO |
|---|---|---|---|---|---|
| 1. NEW | PM2+ observes intake | — | — | — | Owns · triage in 48h |
| 2. BACKLOG | PM2+ frames commercial | — | — | — | Owns · ranks |
| 3. IN REFINEMENT | PM observes | BA2+ chairs · BA1 contributes | L2+ contributes feasibility | L2+ contributes test plan | Attends |
| 4. READY | PM verifies sizing | BA2+ finalizes | L4 Tech Lead signs risk | L2+ confirms test data | Sign-off attended |
| 5. TO DO | PM2+ commits scope | — | Any Dev claims | — | — |
| 6. IN PROGRESS | PM monitors SLA | — | L2+ solo · L1 paired | — | — |
| 7. IN CODE REVIEW | — | — | L3+ gates · L2 co · L1 observes | L2+ pairs on QA-touching PRs | — |
| 8. IN QA (Manual) | — | — | — | L2+ owns · L1 paired | — |
| 8b. IN AUTOMATION | — | — | L3+ for infra changes | L3 SDET-track owns · L2 assists | — |
| 9. READY FOR DEMO | PM2+ pre-checks | BA2+ confirms AC trace | L4 Tech Lead confirms deployability | L3 QualityGatekeeper signs off | — |
| 10. ACCEPTED | PM2+ closes loop in ADO | — | — | — | PO signs off per story |
| 11. DEPLOYED | — | — | L3+ ships · L4 approves risky | — | — |
| 12. DONE | PM2+ verifies SLA | BA2+ closes · updates KH | — | — | — |
Owns this state
Contributes / observes
Not involved
8. A day in the life
Compressed snapshot of a typical day inside a 3-day sprint. Each role has a junior view and a senior view — see how the rhythm changes with level.
PM / Scrum Master
PM1 — Junior
- Shadows planning + standup chair (does not chair yet)
- Runs blocker round-up · escalates to PM2+ in real time
- Drafts status report — PM2+ reviews before client send
- Sits in demo + retro · observes facilitation patterns
PM3 — Senior
- Chairs planning + facilitates retro alone
- Holds commercial conversation with PO on scope change
- Coaches PM1 on hard people moment they faced
- Reviews unit-level process — flags 1 improvement for D3
Business Analyst
BA1 — Junior
- Drafts spec.md content using SpecKit prompts
- Maintains Knowledge House entries
- Attends refinement · BA2+ chairs the call
- Files AC clarifications in ADO comments
BA3 — Senior (Client-Facing)
- Chairs refinement · runs commercial-scope conversation with client mid-sprint
- Co-presents demo · owns the AC trace
- Spots renewal/expansion signal in client feedback log
- Mentors BA1/BA2 on stakeholder mapping after retro
Developer (SE)
L1 — Junior
- Implements well-scoped tasks under L2+ pairing
- Writes unit tests assisted by Copilot · L2+ reviews first
- Observes code review · learns the patterns
- Does not deploy to production solo
L3 — Senior · L4 Tech Lead
- Gates code review for the team · approves merges
- L4: curates Custom Skills + Constitution · approves architectural PRs
- Mentors L1/L2 · pair-programs on tricky tickets
- L3+ ships routine deploys · L4 approves risky deploys
QA / QualityGatekeeper
QA1 — Junior
- Runs prepared test cases · files reproducible bugs
- Paired with QA2+ for any sign-off
- Observes pipeline output · learns Allure + Playwright
- Does not gate stage release
QA3 — QualityGatekeeper
- Signs off stage release · the single human gate
- Owns 5-stage automation pipeline (SDET track) OR test strategy (Analyst track)
- Mentors QA1/QA2 on risk-based test design
- If unavailable: documents handover to backup gatekeeper before sprint starts
9. Reporting channels
Who reports what, where, and how often. Three reports are mandatory on every D3 project. Five more are recommended — adopt them as the project size and client maturity demand.
Mandatory
Required on every project — non-negotiable
Daily
Daily standup
Purpose: surface blockers within 1 day so the sprint doesn't drift unnoticed.
Per sprint
Sprint report
Purpose: contractual evidence of what shipped + input to next-sprint planning.
Weekly
Weekly report
Purpose: surface risks early to Head of D3 + leadership, before they bite.
Recommended
Adopt based on project size, client maturity, and team capacity
Sprint start
★★★
Sprint goal commit
Purpose: align expectations BEFORE work starts. No surprise at demo.
Sprint end
★★★
Demo + acceptance log
Purpose: official artifact for payment milestone. No 'shipped but not accepted' ambiguity.
Weekly
★★
Risk register
Purpose: forward-looking signal — catch risks before they become incidents.
Weekly
★★
AUTO
Quality snapshot
Purpose: catch quality regressions early. Auto-alert if any metric crosses threshold.
Sprint end
★
Sprint retro highlights
Purpose: continuous improvement. Prevent recurring issues sprint after sprint.
Keep reporting lean — 5 principles
- Auto-extract first. If Jira / Linear / CI / test infra already has the data, pull it. Never make humans re-type what tools already track.
- Every report needs an ACTION. If no decision or follow-up comes from a report, kill it. Reports without consequences become theater.
- Short beats long. Daily 30 seconds · weekly ≤ 1 page · per-sprint ≤ 2 pages. Length is the enemy of being read.
- Template beats free-form. Standardize structure across projects. Fill the form, don't write essays. Readers find what they need faster.
- By-product, not extra work. The sprint report = Jira export + 3 lines of PM context. The demo log = recording link + 5 checkboxes. Reports should fall out of doing the work — not be a second job.