Phase Gate Checklist
Print this page. Post it. Every phase gate must pass before the pipeline advances. No exceptions.
Phase −1: Discovery Engine
Purpose: Align stakeholders and establish the domain foundation.
| Checklist Item | Required? | Verified by |
|---|---|---|
| Product Vision Statement agreed | ✅ Mandatory | PO + TL |
| Persona definitions complete (specific actors, not "user") | ✅ Mandatory | PO |
| Feature list prioritized and sliced | ✅ Mandatory | PO + TL |
| MVP canvas complete | ✅ Mandatory | PO |
| Bounded Context map produced | ✅ Mandatory | TL |
| Ubiquitous Language session completed | ✅ Mandatory | TL |
domain-model.md seed created | ✅ Mandatory | TL |
context-map.md created | ✅ Mandatory | TL |
| All discovery artifacts consistent with stated mission | ✅ Mandatory | PO + TL |
Owner: PO + TL jointly
Output blocking next phase: All discovery artifacts must exist. Inconsistencies must be resolved.
Phase 0: Product Intent Approval
Purpose: Formalize strategic purpose and obtain PO sign-off.
| Checklist Item | Required? | Verified by |
|---|---|---|
intent.md contains: mission, capabilities, success metrics, non-goals | ✅ Mandatory | PO |
Entry mode declared (NEW_PRODUCT / IMPROVEMENT / BUG_FIX / REFACTOR) | ✅ Mandatory | PO |
| Bounded context and domain entities listed | ✅ Mandatory | TL |
| PO confidence score ≥ 0.85 | ✅ Mandatory | PO |
| No open questions blocking intent clarity | ✅ Mandatory | PO + TL |
Owner: Product Owner
Output blocking next phase: PO-approved intent.md.
No Phase 1 work begins without an approved intent.md. A "mostly done" intent document will produce hallucinated specs downstream.
Phase 1: AI-Augmented Discovery (Spec Validation Gate)
Purpose: Transform intent into validated, machine-interpretable requirements.
| Checklist Item | Required? | Verified by |
|---|---|---|
All requirements categorized (FEATURE/BUG/IMPROVEMENT) | ✅ Mandatory | Spec Agent → CI |
| All requirements assigned to a slice (MVP/V1/V2) | ✅ Mandatory | Spec Agent |
| EARS syntax compliance | ✅ Mandatory (automated) | CI linter |
All domain terms resolved in domain-model.md | ✅ Mandatory (automated) | Validation Agent |
| Testability score ≥ 0.80 per requirement | ✅ Mandatory | Validation Agent |
| Risk score assigned to every requirement | ✅ Mandatory | Validation Agent |
| QA Agent peer review completed | ✅ Mandatory | QA Agent (Mode C) |
| No open dissent notices against any MUST requirement | ✅ Mandatory | TL |
| Validation Agent confidence ≥ 0.90 | ✅ Mandatory | Validation Agent |
| TL sign-off | ✅ Mandatory | Tech Lead |
Owner: Tech Lead
Output blocking next phase: All MUST requirements in the active slice passed the Spec Validation Gate.
Phase 2: Domain Contracts
Purpose: Produce and approve the schema-compliant domain model.
| Checklist Item | Required? | Verified by |
|---|---|---|
domain-model.md conforms to YAML schema | ✅ Mandatory (automated) | CI gate |
| All entities referenced in Phase 1 requirements exist in model | ✅ Mandatory | Domain Agent |
| All invariants documented per entity | ✅ Mandatory | TL |
| Ubiquitous Language section covers all new terms | ✅ Mandatory | TL |
| No backward-incompatible entity changes without ADR | ✅ Mandatory | TL |
| Domain Agent confidence ≥ 0.85 | ✅ Mandatory | Domain Agent |
| TL sign-off | ✅ Mandatory | Tech Lead |
Owner: Tech Lead
Output blocking next phase: TL-approved, CI-validated domain-model.md.
Phase 3: Architecture Design
Purpose: Synthesize architecture from domain contracts and requirements.
| Checklist Item | Required? | Verified by |
|---|---|---|
design.md contains: overview, component boundaries, DB schema, sequence diagrams | ✅ Mandatory | Design Agent |
| Every design element traceable to ≥1 requirement | ✅ Mandatory | Design Agent |
| Security surface analysis included | ✅ Mandatory | Design Agent |
| Non-functional requirements addressed | ✅ Mandatory | Design Agent |
| ADRs produced for all significant decisions | ✅ Mandatory | Design Agent |
| Design Agent Reasoning Trace produced | ✅ Mandatory | Design Agent |
| TL has acknowledged the Reasoning Trace | ✅ Mandatory | Tech Lead |
| Design Agent confidence ≥ 0.80 | ✅ Mandatory | Design Agent |
TL sign-off on design.md | ✅ Mandatory | Tech Lead |
Owner: Tech Lead
Output blocking next phase: TL-approved design.md with acknowledged Reasoning Trace.
Signing off on the artifact without reviewing the Reasoning Trace is rubber-stamping. The Reasoning Trace is part of the gate, not an optional appendix.
Phase 4: Agentic Implementation
Purpose: Generate spec-compliant code through parallel execution waves.
| Checklist Item | Required? | Verified by |
|---|---|---|
tasks.md wave plan approved | ✅ Mandatory | TL |
| All wave tasks trace to a spec requirement | ✅ Mandatory | Task Planning Agent |
No UNDOCUMENTED_BEHAVIOR flagged | ✅ Mandatory | Implementation Agent |
| All TDD tests pass (RED → GREEN → REFACTOR) | ✅ Mandatory | QA Agent |
| Spec Coverage = 100% for this slice | ✅ Mandatory | QA Agent (Mode B) |
| Test coverage ≥ minimum threshold | ✅ Mandatory | QA Agent (Mode B) |
| Implementation Agent Reasoning Traces acknowledged | ✅ Mandatory | TL + Engineer |
| Human code review completed (when confidence < 0.75) | ✅ Conditional | Engineer |
Owner: Tech Lead + Engineers
Output blocking next phase: All tasks complete, TDD green, QA gate passed, Reasoning Traces acknowledged.
Phase 5: Autonomous CI/CD
Purpose: Automated validation of the full integrated system.
| Checklist Item | Required? | Verified by |
|---|---|---|
| All CI pipeline stages green | ✅ Mandatory | DevOps Agent |
| Security Agent scan completed | ✅ Mandatory | Security Agent |
| Security Agent confidence ≥ 0.95 | ✅ Mandatory | Security Agent |
| Security scan artifact attached to pipeline | ✅ Mandatory | CI |
| No force-bypass of any CI gate | ✅ Mandatory | CI |
| All specification coverage checks passed | ✅ Mandatory | CI |
| TL sign-off (if any pipeline failure occurred) | ✅ Conditional | Tech Lead |
Owner: DevOps Agent + Tech Lead
Output blocking next phase: All CI gates green; no force-bypass; security scan artifact attached.
Phase 6: Observability
Purpose: Ensure the deployed system is measurable and errors are detectable.
| Checklist Item | Required? | Verified by |
|---|---|---|
telemetry-plan.md created | ✅ Mandatory | Observability Agent |
| API latency telemetry emitting (p50/p95/p99) | ✅ Mandatory | Observability Agent |
| Error rate by type emitting | ✅ Mandatory | Observability Agent |
| Transaction success/failure counts emitting | ✅ Mandatory | Observability Agent |
| Business event counts mapped to domain events | ✅ Mandatory | Observability Agent |
| All new components confirmed emitting | ✅ Mandatory | Observability Agent |
| TL sign-off | ✅ Mandatory | Tech Lead |
Owner: Tech Lead
Output blocking next phase: Observability Agent confirms telemetry emission from all new components.
Phase 7: Production Learning Loop
Purpose: Close the loop between production and specifications.
| Checklist Item | Required? | Verified by |
|---|---|---|
| Knowledge Agent failure pattern analysis completed | ✅ Mandatory | Knowledge Agent |
| Any steering proposals include: pattern data (≥3 incidents), rule text, predicted outcome, rollback plan | ✅ Conditional | Knowledge Agent |
| TL + Engineer have reviewed all DRAFT steering proposals | ✅ Conditional | TL + Engineer |
Approved proposals merged to .asdd/steering/ | ✅ Conditional | TL |
| Self-Healing PRs (if any): TL + Engineer approved, CI passed, rollback documented | ✅ Conditional | TL + Engineer |
lessons-learned.md updated | ✅ Mandatory | TL |
| All open Self-Healing PRs reviewed | ✅ Mandatory | TL |
Owner: Tech Lead + Product Owner
Gate closes when: Knowledge Agent proposals reviewed; approved updates applied; no open Self-Healing PRs pending review.
Agent failure quick reference
| Situation | Immediate Action | Owner |
|---|---|---|
| Agent confidence below threshold | Block phase gate; log; escalate to TL | Automated + TL |
| CCS < 0.65 | Halt entire slice pipeline; full human review | TL |
| Conflicting agent outputs | Pause pipeline; Knowledge Agent arbitrates; TL resolves | TL |
| Agent timeout or crash | Retry once; if still failed, escalate to TL | Automated + TL |
| Dissent notice filed | Log in dissent log; TL resolves within SLA | TL |
| Self-Healing PR proposed | Mandatory TL + Engineer review; CI must pass | TL + Engineer |
| Self-Healing PR reverted | Knowledge Agent flags steering update for re-analysis | Knowledge Agent + TL |
| Discovery Re-entry triggered | Slice resets to Phase −1; PO + TL notified | PO + TL |