Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Analyze and optimize cloud costs across AWS, Azure, GCP with rightsizing, reserved instances, waste detection, and FinOps best practices.
.claude/skills/williamzujkowski-cloud-cost-optimization-analyzer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 19% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 89% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 270% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 135% | 0% |
Primary trigger conditions:
When NOT to use this skill:
Value proposition: Identifies 30-40% of typical cloud waste through systematic analysis of rightsizing, commitment discounts, and idle resources using FinOps principles.
Required inputs validation:
pythonNOW_ET = "2025-10-25T22:42:30-04:00" assert cloud_provider in ["aws", "azure", "gcp", "multi"], "Invalid cloud provider" assert cost_data_source is not None, "Cost data source required" assert time_range in ["7d", "30d", "90d"] or matches_custom_format(time_range) assert len(optimization_targets) > 0, "At least one optimization target required" # Data freshness check if cost_data_age > 24h: warn("Cost data is stale; recommendations may be outdated") # Minimum data volume check if time_range == "7d" and total_resources < 10: suggest("Use 30d+ range for better trend analysis")
Authority checks:
ce:GetCostAndUsage permissionbilling.accounts.getSpendingInformation permissionSource citations (accessed 2025-10-25T22:42:30-04:00):
Goal: Identify top 3 cost optimization opportunities in <5 minutes.
Steps:
Token budget checkpoint: ~1.5k tokens for API calls, analysis, and output formatting.
Goal: Generate detailed, actionable cost optimization plan with quantified savings.
Extends T1 with:
(current_price - recommended_price) * hours_per_monthupfront_cost / monthly_savings(current_storage_cost - optimized_cost) * TB_storedAuthority sources (accessed 2025-10-25T22:42:30-04:00):
Output: JSON report with sections: cost_summary, quick_wins (T1), rightsizing_recommendations, commitment_recommendations, storage_optimization, network_optimization, anomalies, finops_maturity_score.
Token budget checkpoint: ~5k tokens (includes T1 + extended analysis + detailed output).
Reserved for future enhancements: predictive cost forecasting, ML-based anomaly detection, multi-account/org-wide consolidation, custom FinOps policies.
When to abort:
Ambiguity thresholds:
Prioritization logic:
savings_per_month / implementation_effortFinOps principle application (accessed 2025-10-25T22:42:30-04:00):
Per FinOps Foundation principles (https://www.finops.org/framework/principles/):
Schema (JSON):
json{ "cost_analysis_report": { "period": "2025-09-25 to 2025-10-25", "cloud_provider": "aws", "total_spend": 47500.32, "waste_identified": 14250.10, "savings_potential": { "monthly": 12800.00, "annual": 153600.00, "percentage": 26.9 } }, "quick_wins": [ { "category": "unused_resources", "description": "Delete 12 unattached EBS volumes", "monthly_savings": 360.00, "implementation_effort": "low", "risk_level": "none" } ], "rightsizing_recommendations": [ { "resource_id": "i-0a1b2c3d4e5f6g7h8", "resource_type": "ec2_instance", "current_type": "m5.2xlarge", "recommended_type": "m5.xlarge", "utilization_avg": 18.5, "monthly_savings": 150.00, "confidence": "high" } ], "commitment_recommendations": [ { "service": "ec2_compute", "plan_type": "compute_savings_plan", "term": "3_year", "upfront": "partial", "monthly_commitment": 5000.00, "monthly_savings": 1200.00, "break_even_months": 4.2 } ], "waste_inventory": [ { "resource_id": "vol-0123456789abcdef", "type": "ebs_volume_unattached", "idle_days": 45, "monthly_cost": 30.00 } ], "action_plan": [ { "priority": 1, "action": "Delete unused resources", "impact": "high", "effort": "low", "items_count": 27, "monthly_savings": 810.00 } ] }
Required fields: cost_analysis_report (with total_spend, savings_potential), action_plan (prioritized).
Optional fields: rightsizing_recommendations, commitment_recommendations (only if applicable).
yaml# Example: AWS cost optimization for over-provisioned workload input: cloud_provider: aws cost_data_source: cost_explorer_api optimization_targets: [compute, storage, network] budget_constraints: monthly_budget: 50000 alert_threshold: 0.90 time_range: 90d output: cost_analysis_report: total_spend: 47500.32 waste_identified: 14250.10 savings_potential: monthly: 12800 annual: 153600 percentage: 26.9 quick_wins: - category: unused_resources description: Delete 12 unattached EBS volumes monthly_savings: 360 rightsizing_recommendations: - resource_id: i-0a1b2c3d4e5f current_type: m5.2xlarge recommended_type: m5.xlarge utilization_avg: 18.5 monthly_savings: 150
Token budgets (enforced):
Accuracy requirements:
Safety constraints:
Auditability:
Determinism:
Official cloud provider documentation:
FinOps Foundation resources:
Third-party cost optimization guides:
Related skills:
cloud-aws-architect: For architecture-level cost optimization during design phasedevops-pipeline-architect: For CI/CD cost optimization (ephemeral environments)cloud-native-deployment-orchestrator: For Kubernetes cost optimization (cluster rightsizing)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 21,511 | 10,406 | -52% | 1 | 1 | 0% | 5,564 | 6,639 | +19% | 0 | 0 | — |
case-02 | fail→pass | 13,212 | 9,995 | -24% | 1 | 1 | 0% | 3,375 | 6,374 | +89% | 0 | 0 | — |
case-07 | pass→pass | 5,334 | 6,815 | +28% | 1 | 1 | 0% | 1,138 | 5,242 | +361% | 0 | 0 | — |
case-03 | fail→pass | 25,939 | 11,782 | -55% | 1 | 1 | 0% | 6,206 | 6,738 | +9% | 0 | 0 | — |
case-04 | fail→fail | 14,531 | 12,511 | -14% | 1 | 1 | 0% | 2,653 | 6,331 | +139% | 0 | 0 | — |
case-05 | fail→pass | 7,151 | 4,648 | -35% | 1 | 1 | 0% | 1,276 | 4,725 | +270% | 0 | 0 | — |
case-06 | pass→pass | 10,908 | 3,985 | -63% | 1 | 1 | 0% | 2,186 | 4,455 | +104% | 0 | 0 | — |
case-08 | fail→fail | 6,661 | 7,900 | +19% | 1 | 1 | 0% | 1,450 | 5,558 | +283% | 0 | 0 | — |
case-09 | pass→pass | 11,154 | 7,598 | -32% | 1 | 1 | 0% | 2,076 | 5,058 | +144% | 0 | 0 | — |
case-10 | fail→pass | 11,422 | 6,888 | -40% | 1 | 1 | 0% | 2,091 | 4,919 | +135% | 0 | 0 | — |
case-11 | pass→pass | 8,320 | 5,147 | -38% | 1 | 1 | 0% | 1,390 | 4,634 | +233% | 0 | 0 | — |
case-12 | pass→pass | 8,997 | 5,057 | -44% | 1 | 1 | 0% | 1,547 | 4,769 | +208% | 0 | 0 | — |
case-13 | fail→pass | 10,547 | 3,911 | -63% | 1 | 1 | 0% | 1,732 | 4,488 | +159% | 0 | 0 | — |
case-14 | pass→pass | 9,702 | 4,814 | -50% | 1 | 1 | 0% | 1,693 | 4,672 | +176% | 0 | 0 | — |
case-15 | fail→pass | 11,867 | 4,055 | -66% | 1 | 1 | 0% | 1,974 | 4,533 | +130% | 0 | 0 | — |
case-16 | pass→pass | 9,923 | 10,066 | +1% | 1 | 1 | 0% | 2,186 | 5,731 | +162% | 0 | 0 | — |
case-17 | pass→pass | 4,865 | 2,901 | -40% | 1 | 1 | 0% | 881 | 4,305 | +389% | 0 | 0 | — |
case-18 | pass→pass | 8,361 | 8,731 | +4% | 1 | 1 | 0% | 1,459 | 5,494 | +277% | 0 | 0 | — |
case-19 | pass→pass | 10,340 | 7,346 | -29% | 1 | 1 | 0% | 1,617 | 4,965 | +207% | 0 | 0 | — |
case-20 | fail→pass | 7,170 | 5,907 | -18% | 1 | 1 | 0% | 1,226 | 4,863 | +297% | 0 | 0 | — |
case-21 | pass→pass | 13,687 | 7,318 | -47% | 1 | 1 | 0% | 2,170 | 5,087 | +134% | 0 | 0 | — |
case-22 | pass→pass | 12,454 | 6,514 | -48% | 1 | 1 | 0% | 2,319 | 5,001 | +116% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +36 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.