TheDocumentation Index
Fetch the complete documentation index at: https://mintlify.com/aws-samples/sample-well-architected-skills-and-steering/llms.txt
Use this file to discover all available pages before exploring further.
reliability-improvement-plan skill performs a focused assessment of your workload’s reliability posture. It analyzes IaC, scaling configurations, and resilience patterns in your codebase to identify single points of failure, assess recovery capabilities, and produce a prioritized remediation plan tied to specific code evidence.
Use
reliability-improvement-plan when you need a dedicated reliability review, SPOF analysis, DR assessment, or failover configuration check. For a multi-pillar review that includes reliability alongside security, cost, and other pillars, use wa-review instead.What the Agent Analyzes
The skill runs a structured discovery across five reliability domains before evaluating against WA Framework questions.Fault Tolerance Discovery
The agent examines infrastructure for single points of failure:
- Compute deployments (AZ distribution, instance count, ASG configs)
- Database configurations (Multi-AZ, read replicas, cluster topology)
- Cache configurations (cluster mode, replica counts, failover)
- Load balancer configurations (cross-zone, health checks, target groups)
- NAT Gateway placement (single vs per-AZ)
- DNS configurations (Route 53 health checks, failover routing)
- Queue and messaging configs (DLQ, redrive policies)
- Storage redundancy (S3 replication, EBS snapshots, EFS)
- Single-AZ database deployments for production workloads
- Compute without auto-scaling (fixed instance count)
- No health checks on load-balanced targets
- Single NAT Gateway serving multiple AZs
- Stateful services without replication
- Missing DLQ on async invocations (Lambda, SQS, EventBridge)
- No circuit breaker or timeout on external service calls
Recovery Capability Discovery
The agent analyzes backup and recovery configurations for every stateful resource:
- AWS Backup plans and rules
- RDS automated backup settings (retention, PITR)
- S3 versioning and replication rules
- DynamoDB PITR and backup settings
- EBS snapshot configurations
- Cross-region replication rules
- Disaster recovery configurations (pilot light, warm standby resources)
- Stateful resources with no backup configuration
- Backup retention < 7 days for production data
- No cross-region backup for critical data
- No evidence of recovery testing
Scaling and Capacity Discovery
The agent examines scaling and capacity configurations:
- Auto Scaling Group configurations (min, max, desired, scaling policies)
- ECS service scaling (target tracking, step scaling)
- Lambda concurrency settings (reserved, provisioned)
- DynamoDB capacity mode (on-demand vs provisioned, auto-scaling)
- SQS/Kinesis throughput configurations
- API Gateway throttling settings
- Service quota usage and alarms
- Compute without auto-scaling policies
- ASG with min = max (no scaling headroom)
- No service quota alarms
- Lambda without reserved concurrency on critical functions
- No load shedding or throttling for overload scenarios
Resilience Pattern Discovery
The agent analyzes application code for resilience patterns:
- Retry configurations (SDK clients, custom retry logic)
- Timeout settings (HTTP clients, database connections, Lambda timeout)
- Circuit breaker implementations
- Fallback logic and graceful degradation patterns
- Idempotency handling (idempotency keys, deduplication)
- Health check endpoint implementations
- Connection pooling configurations
- External service calls without timeouts
- No retry logic on SDK clients
- Missing idempotency on event-driven processing
- Health checks that don’t verify actual functionality (shallow checks)
- Lambda timeout ≥ API Gateway timeout (will always timeout to caller)
Change Management Discovery
The agent reviews deployment safety configurations:
- Deployment strategies (canary, blue/green, rolling, all-at-once)
- Health check gating on deployments
- Automated rollback configurations (alarm-based)
- Database migration strategies (backward-compatible, blue/green schema)
- Feature flag usage
- All-at-once deployment to production
- No automated rollback on health check failure
- Database migrations that aren’t backward-compatible
- No pre-production environment mirroring production topology
WA Framework Coverage: REL 1–13
After discovery, the agent evaluates your workload against all 13 Reliability pillar questions.| Question | Focus Area |
|---|---|
| REL 1 | Service quotas and constraints — quota alarms, SDK retry configs, throttling |
| REL 2 | Network topology planning — subnet definitions, AZ distribution, NAT redundancy |
| REL 3 | Adapting to demand — ASG configs, scaling policies, Lambda concurrency, DynamoDB capacity mode |
| REL 4 | Distributed system failure prevention — retry logic, timeout configs, SQS decoupling, idempotency |
| REL 5 | Failure mitigation — circuit breakers, fallback paths, bulkhead patterns, load shedding |
| REL 6 | Workload monitoring — health check endpoints, alarm definitions, composite alarms, dashboards |
| REL 7 | Demand adaptation — scaling policy metrics, scheduled scaling, predictive scaling |
| REL 8 | Change implementation — deployment configs, health check gating, rollback trigger alarms |
| REL 9 | Data backup — AWS Backup plans, PITR settings, replication rules, snapshot configs |
| REL 10 | Fault isolation — AZ distribution, cell-based patterns, shuffle sharding, isolation boundaries |
| REL 11 | Component failure withstand — multi-AZ configs, failover policies, stateless design, health-based routing |
| REL 12 | Reliability testing — FIS experiments, failure injection code, game day runbooks, DR test scripts |
| REL 13 | Disaster recovery planning — cross-region resources, DR automation, backup restore procedures, RTO/RPO docs |
Output Format
The skill produces a structured reliability improvement plan including:- Reliability Scorecard — 1–5 score across six domains (Fault Tolerance, Recovery & Backup, Scaling & Capacity, Resilience Patterns, Change Management, Testing & Validation)
- Single Points of Failure table — component, evidence (file:line), failure impact, current mitigation, risk level
- Critical and High Risk Findings — with domain, title, description, evidence, impact assessment, recommendation, effort, and relevant AWS services
- Medium and Low Risk Findings — in condensed format
- Prioritized Remediation Plan — Quick Wins (< 1 week), Foundation (1–4 weeks), Strategic (1–3 months)
- Testing Plan — AZ failover, database failover, load test, backup restore, deployment rollback with FIS and frequency recommendations
How to Invoke
Example Output: Single Points of Failure Table
Matching Expectations to Availability Targets
The agent calibrates its findings based on your stated availability target. A 99.9% SLA does not require multi-region architecture — single-region multi-AZ is sufficient. A 99.99% SLA does require multi-region active-active or active-passive DR. Specify your target when invoking the skill for appropriately scoped recommendations.
| Availability Target | Typical Requirements |
|---|---|
| 99.9% (~8.7h/year downtime) | Multi-AZ compute and databases, auto-scaling, automated rollback |
| 99.95% (~4.4h/year downtime) | Above + chaos engineering, DR testing, connection resilience |
| 99.99% (~52m/year downtime) | Above + multi-region active-passive DR, global load balancing |
| 99.999% (~5m/year downtime) | Multi-region active-active, cell-based architecture, extensive FIS testing |
Benchmark Results
Evaluated with Claude Opus 4.8, 16K output tokens, paired comparison (same prompt with and without skill):| Baseline | With Skill | Delta |
|---|---|---|
| 95% | 100% | +5% |