Rollout Skill

Plan and execute safe, progressive feature rollouts.

Quick Start

bash

# Plan rollout for feature
/wicked-delivery:rollout feature-new-dashboard

# Plan with specific strategy
/wicked-delivery:rollout feature-checkout --strategy canary

# Discover deployment tools
/wicked-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

Overall Risk

code

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.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

markdown

## Rollout Plan: {Feature Name}

### Risk Assessment
- **Overall Risk**: {LOW | MEDIUM | HIGH}
- **User Impact**: {%} of users
- **Revenue Impact**: {LOW | MEDIUM | HIGH}
- **Reversibility**: {HIGH | MEDIUM | LOW}

### Strategy: {Progressive | Canary | Big Bang}
**Duration**: {weeks}
**Feature Flag**: {flag_name}

### Rollout Stages

| Stage | Traffic | Duration | Success Criteria | Auto Rollback |
|-------|---------|----------|------------------|---------------|
| 1 | 1% | 24h | Error < 0.1% | Error > 0.5% |
| 2 | 10% | 3d | Conversion >= baseline | Conv < 95% |

### Monitoring
**Dashboard**: {url}
**Key Metrics**: {list}
**Alerts**: WARNING ({conditions}), CRITICAL ({conditions})

### Rollback Plan
**Automatic**: {triggers}
**Manual**: {criteria}
**Procedure**: 1. Set flag to 0%, 2. Verify, 3. Monitor, 4. Notify

### Communication
**Pre-Launch**: {stakeholder checklist}
**During**: {update frequency}
**Post-Launch**: {summary and learnings}

### Stage Gate Checklist
- [ ] Feature flag configured
- [ ] Monitoring live
- [ ] Alerts tested
- [ ] On-call rotation set
- [ ] Rollback tested

### Next Steps
1. {Action 1}
2. {Action 2}

Capability Discovery

Discovers deployment tools automatically via capability detection:

Capabilities needed:

•feature-flags: Feature toggle and flag management
•deployment: Progressive rollout and canary deployment tools
•monitoring: 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 wicked-kanban: Tracks rollout progress per stage With wicked-qe: Uses test scenarios for validation With wicked-mem: Recalls past rollout learnings With wicked-delivery:analyze: Uses experiment results for confidence