| name | rollout |
| description | Plan and coordinate progressive feature rollouts.
Risk assessment, canary deployments, feature flag management,
rollback procedures. Discovers deployment tools via capabilities.
Use when: "roll out feature", "progressive rollout", "canary deployment",
"feature flag", "rollback plan", "launch feature", "deploy gradually"
|
| phase_relevance | ["build","review","operate"] |
| archetype_relevance | ["*"] |
Rollout Skill
Plan and execute safe, progressive feature rollouts.
Quick Start
/wicked-garden:delivery:rollout feature-new-dashboard
/wicked-garden:delivery:rollout feature-checkout --strategy canary
/wicked-garden:delivery:rollout --discover
What This Skill Does
- Assesses rollout risk
- Selects appropriate rollout strategy
- Defines progressive stages
- Establishes success and rollback criteria
- Creates monitoring and communication plan
Risk Assessment
Risk Factors
User impact: <1% (LOW) | 1-25% (MEDIUM) | >25% (HIGH)
Revenue impact: Non-revenue (LOW) | Funnel (MEDIUM) | Direct payment (HIGH)
System criticality: Nice-to-have (LOW) | Important (MEDIUM) | Mission-critical (HIGH)
Reversibility: Flag toggle (HIGH) | Schema change (MEDIUM) | Data migration (LOW)
Overall Risk
Risk = MAX(user, revenue, criticality) + reversibility_penalty
Rollout Strategies
| Risk | Strategy | Timeline | Stages |
|---|
| LOW | Big Bang | Immediate | 0% → 100% |
| MEDIUM | Progressive | 1-2 weeks | 0% → 10% → 25% → 50% → 100% |
| HIGH | Canary | 4-6 weeks | 0% → 1% → 5% → 10% → 25% → 50% → 100% |
See strategies-core.md and strategies-operations.md for detailed guidance.
Stage Definition
For each stage:
- Traffic percentage: 1%, 10%, 25%, 50%, 100%
- Duration: 24h to 1 week
- Success criteria: Error rate, conversion, performance
- Rollback criteria: Automatic and manual triggers
Success & Rollback Criteria
Success (all must pass):
- Primary metric stable or improved
- Error rate within bounds
- Performance acceptable
- No critical bugs
Automatic rollback:
- Error rate >2x baseline
- Performance >2x latency
- Availability <99.5%
Manual rollback:
- Revenue impact detected
- Security vulnerability
- Regulatory issue
Monitoring Plan
Business: Primary metric, revenue, retention
Technical: Error rate, latency (p50/p95/p99), throughput
User: Support tickets, feedback, crashes
Alerts:
- WARNING: 1.5x baseline
- CRITICAL: 2x baseline or 5% error rate
Communication Plan
Stakeholders: Engineering (technical), Product (timeline), Support (FAQs), Leadership (risk)
Timeline:
- T-2 days: Pre-rollout notification
- T-0: Rollout begins
- T+checkpoints: Progress updates
- T+complete: Completion summary
Output Format
The skill emits a markdown rollout plan with sections: Risk Assessment, Strategy, Rollout Stages (table), Monitoring, Rollback Plan, Communication, Stage Gate Checklist, Next Steps.
Full template with substitutable placeholders: refs/output-template.md.
Capability Discovery
Discovers deployment tools automatically via capability detection:
Capabilities needed:
feature-flags: Feature toggle and flag managementdeployment: Progressive rollout and canary deployment toolsmonitoring: Metrics, dashboards, and alerting
Discovery methods:
- CLI tools (container orchestration, feature flag CLIs, IaC tools)
- Configuration files (deployment manifests, infrastructure configs)
- Environment variables (API keys, endpoints)
- SDK detection (package dependencies)
Asks "Do I have deployment capability?" not "Do I have Kubernetes?"
Gracefully degrades to manual procedures when capabilities unavailable.
Integration
With native tasks: Tracks rollout progress per stage via TaskCreate/TaskUpdate with metadata.event_type="task"
With qe: Uses test scenarios for validation
With wicked-brain:memory: Recalls past rollout learnings
With wicked-garden:delivery:experiment: Uses experiment results for confidence
See Also
