Skip to main content

Product Management Lifecycle

xOps uses a 6-phase PM lifecycle anchored to the ADLC framework. The product-owner agent leads each phase; HITL approves scope and ships code.

How to Start: .specify/ Workflow + Product Skills

Two complementary tools work together across the 6 phases:

.specify/ WorkflowProduct Skills & Ceremonies
What6 sequential commands that produce spec → plan → tasks → codeDiscovery skills before spec, ceremony commands after build
WhenEvery sprint — the backbone processBefore /speckit.specify (discovery) and after /speckit.implement (ceremonies)
WhyStructured governance prevents scope creep and NATOSkills surface real problems; ceremonies measure real progress

CloudOps (runbooks PyPI package)

Phase.specify/ CommandProduct Skill / Ceremony
Discover/product:pr-faq CloudOps-S2lean-canvas, competitive-battlecard
Specify/speckit.specify — CLI sync validator specopportunity-solution-tree for 3-way sync problem
Plan/speckit.plan — architecture + agent delegation
Build/speckit.implement — python-engineer executes
Measure/metrics:daily-standup, /metrics:sprint-review
Iterate/metrics:sprint-retro, /speckit.retrospective

xOps (F2T2EA operational cycle)

Phase.specify/ CommandProduct Skill / Ceremony
Discover/product:pr-faq xOps-S2lean-canvas, customer-journey-map
Specify/speckit.specify — F2T2EA cycle specnorth-star-metric for find-to-fix time
Plan/speckit.plan — 6-layer sovereign stack
Build/speckit.implement — multi-agent execution
Measure/metrics:daily-standup, /metrics:sprint-review
Iterate/metrics:sprint-retro, /speckit.retrospective

Golden path: Run the .specify/ commands top-to-bottom for each sprint. Use product skills in Discover phase, ceremonies in Measure/Iterate phases. See Pillar 1: Product Management for the full component map.

6 Phases

Discover

Who: HITL steers direction, product-owner agent validates business fit

What: Find problems worth solving for 1 Cloud Ops Manager operating a multi-account AWS environment

When: Sprint 0 — before first code sprint

Why: Without discovery, you build solutions nobody needs. F2T2EA was discovered because manual multi-account operations took hours and missed resources across dozens of accounts.

Evidence of Value: 13 INVEST stories validated (PO=84, CA=82, QA=87). Clear north-star for 8-session sprint.

Commands: /product:pr-faq xops, lean-canvas, opportunity-solution-tree

Output: PR/FAQ + 13 INVEST stories + F2T2EA north-star


Specify

Who: product-owner agent drafts, HITL approves scope

What: Define F2T2EA operational cycle — 6 phases, 7 components, governance gates G0–G6

When: Sprint 0 — after discovery validates the domain

Why: Spec-Driven Development makes specs executable. The xOps spec defined 13 stories with acceptance criteria that agents could independently implement.

Evidence of Value: Precise scope — 4 stories in CloudOps-S1, 9 in xOps-S2. Zero scope creep.

Commands: /speckit.specify, /speckit.clarify

Output: spec.md + 13 INVEST stories with acceptance criteria


Plan

Who: cloud-architect agent designs 6-layer architecture, product-owner validates business alignment

What: Design L1 Identity through L6 Interface. Map agents to tasks. Constitution check against 7 ADLC principles.

When: Sprint planning — before first Build sprint

Why: Architecture decisions made here (ECS vs EKS, SQLite vs RDS, Claude API vs Bedrock) shape cost and feasibility for 2026–2030.

Evidence of Value: 6-layer architecture with cost estimates ($110–460/mo). 2 Terraform modules published (M1, M2).

Commands: /speckit.plan, /speckit.tasks

Output: plan.md + tasks.md + 6-layer architecture + agent delegation matrix


Build

Who: AI agents (python-engineer, infrastructure-engineer) build and execute. HITL reviews and ships trusted output.

What: 8-session sprint execution with PDCA quality gates. Specialist agents execute tasks, orchestrator coordinates.

When: CloudOps-S1 sprint (14 days) + xOps-S2 sprint (planned)

Why: Agents prepare, humans decide. Governed execution with evidence prevents drift. Each session produces code, tests, and evidence.

Evidence of Value: 1,430 LOC new code across 10 modules. 98.3% cross-validation accuracy. v1.3.10 published.

Commands: /speckit.implement, specialist agents

Output: Working software + test evidence + PDCA rounds


Ceremony

Who: observability-engineer collects metrics, HITL reads the dashboard and decides what to improve

What: Daily standup, sprint review, retrospective. DORA metrics. Agent consensus scoring (PO+CA+QA).

When: Daily (standup), end of sprint (review + retro)

Why: What gets measured gets managed. Ceremonies create the evidence trail that prevents NATO and ensures continuous improvement.

Evidence of Value: DORA 3/4 GREEN. 100% BV delivered (1472/1472). Sprint health 90/100.

Commands: /metrics:daily-standup, /metrics:sprint-review

Output: DORA dashboard + sprint evidence + ceremony artifacts


Iterate

Who: Full agent team reflects. HITL decides what to change in the process.

What: Retrospective with 4L review. Extract patterns. Update constitution. Improve the golden path itself.

When: End of each sprint + after major incidents

Why: The process that builds products must itself improve continuously. Improvement compounds across sprints.

Evidence of Value: 22 sessions of lessons learned. 64 anti-patterns cataloged. Constitution v2.0.0.

Commands: /metrics:sprint-retro, /speckit.retrospective

Output: Improvement actions + updated constitution + PDCA evidence


Skills Used Per Phase

PhaseSkillsCount
Discoverylean-canvas, opportunity-solution-tree, identify-assumptions-new, pr-faq4
Strategyproduct-vision, product-strategy, competitive-battlecard, north-star-metric4
Execution/speckit.specify, /speckit.plan, /speckit.tasks, /speckit.implement4
Ceremony/metrics:daily-standup, /metrics:sprint-review, /metrics:sprint-retro3

DORA Definitions

MetricMeasuresTarget
Deploy Frequency (DF)Throughput — merge-to-main deploys per sprint≥2/sprint
Lead Time (LT)Speed — time from first commit to merge<2 days
Change Failure Rate (CFR)Stability — % of deploys causing rollback or hotfix<5%
MTTRResilience — time to restore service after failure<1 hour