Cost

Cost — Cloud Cost Optimization

Activate When

• User invokes /godmode:cost
• User says "reduce cloud costs," "optimize spending," "why is our bill so high?"
• User asks about right-sizing, reserved instances, or spot pricing
• Godmode orchestrator detects infrastructure cost concerns
• After /godmode:infra provisions resources that need cost governance

Workflow

Step 1: Inventory Cloud Resources

Discover all provisioned resources and their current costs:

COST INVENTORY:
Provider: 
Account(s): 
Region(s): 
Time period: 

Resource categories:
  Compute: 
  Storage: 
  Database: 
  Network: 
  Containers: 
  Serverless: 
  Other: 

Step 2: Utilization Analysis

Measure actual usage versus provisioned capacity:

Compute Utilization

COMPUTE UTILIZATION:
| Instance | Type | Avg CPU | Avg Mem | Verdict |
|--|--|--|--|--|
|  | m5.2xl | 12% | 25% | OVERSIZE |
|  | t3.micro | 89% | 92% | UNDERSIZE |
|  | c5.large | 45% | 60% | OK |
|  | m5.xl | 3% | 8% | IDLE |

Thresholds:
  IDLE:  80% CPU or > 85% memory sustained
  OK: within healthy range

Storage Utilization

STORAGE UTILIZATION:
| Bucket/Volume | Size | Access | Last Hit | Verdict |
|--|--:|--|--|--|
|  | 2.3 TB | Frequent | Today | OK |
|  | 500 GB | None | 90d ago | ARCHIVE |
|  | 1 TB | None | Never | DELETE |
|  | 200 GB | N/A | 180d ago | DELETE |

Database Utilization

DATABASE UTILIZATION:
| Instance | Type | Avg CPU | Storage | Verdict |
|--|--|--:|--:|--|
|  | db.r5.xl | 35% | 40% | OK |
|  | db.r5.xl | 5% | 10% | OVERSIZE |
|  | db.m5.lg | 2% | 5% | SCHEDULE |

Step 3: Waste Detection

Identify resources that cost money but provide no value:

WASTE DETECTION:
| Category | Count | Monthly Cost | Action |
|--|--|--|--|
| Unattached EBS vols | 12 | $340 | DELETE |
| Old snapshots (>90d) | 45 | $180 | DELETE |
| Idle load balancers | 3 | $75 | DELETE |
| Unused Elastic IPs | 8 | $29 | RELEASE |
| Orphaned ENIs | 5 | $0 | CLEANUP |
| Dev envs running 24/7 | 4 | $1,200 | SCHEDULE |
| Oversized instances | 6 | $2,400 waste | RESIZE |
| Stale DNS records | 15 | $0 | CLEANUP |

Total identifiable waste: $4,224/month ($50,688/year)

Step 4: Right-Sizing Recommendations

For each oversized or undersized resource, recommend the optimal size:

RIGHT-SIZING RECOMMENDATIONS:
| Resource | Current | Recommended | Monthly Savings |
|--|--|--|--|
|  | m5.2xlarge | m5.large | $180 (65% less) |
|  | db.r5.xl | db.t3.medium | $420 (78% less) |
|  | r6g.xlarge | r6g.large | $95 (50% less) |
|  | m5.xlarge | TERMINATE | $140 (100% saved) |

Basis: 14-day P95 utilization data.
Risk: LOW — all recommendations leave 40%+ headroom above P95.

Step 5: Pricing Optimization

Recommend pricing model changes for stable workloads:

Reserved Instances / Savings Plans

RESERVATION RECOMMENDATIONS:
| Resource | On-Demand | Reserved(1y) | Savings |
|--|--|--|--|
| Prod compute (6x) | $2,400/mo | $1,560/mo | $840/mo (35%) |
| Prod database (2x) | $1,200/mo | $780/mo | $420/mo (35%) |
| Prod cache (2x) | $380/mo | $247/mo | $133/mo (35%) |

Prerequisites: Workload must have run for 3+ months with stable utilization.
Commitment: 1-year, no upfront (lowest risk).

Spot / Preemptible Instances

SPOT CANDIDATES:
- CI/CD runners:  instances, tolerant of interruption → 60-70% savings
- Batch processing:  instances, can retry → 60-70% savings
- Dev environments:  instances, non-critical → 60-70% savings

NOT spot-eligible: production web servers, databases, stateful services.

Step 6: Cost Allocation & Tagging

Verify all resources are tagged for cost attribution:

TAGGING AUDIT:
| Required Tag | Coverage | Missing | Action |
|--|--|--|--|
| team | 72% | 45 res | TAG |
| environment | 85% | 24 res | TAG |
| project | 60% | 64 res | TAG |
| cost-center | 45% | 88 res | TAG |
| owner | 55% | 72 res | TAG |

Recommended tagging policy:
  REQUIRED: team, environment, project, cost-center
  RECOMMENDED: owner, created-by, expiry-date
  ENFORCED VIA: AWS Config rules / GCP Organization Policy / Azure Policy

Step 7: Budget Alerts

Set up proactive cost monitoring:

BUDGET ALERT CONFIGURATION:
| Budget | Monthly Limit | Alert at | Notify |
|--|--|--|--|
| Total account | $15,000 | 50/80/100% | #finops, PagerDuty |
| Production | $10,000 | 80/100% | #infra |
| Development | $3,000 | 80/100% | #dev-team |
| Per-service | varies | 100/120% | service owner |

Anomaly detection:
  - Alert if daily spend exceeds 2x rolling 7-day average
  - Alert if any single resource exceeds $500/day
  - Weekly cost digest to #finops channel

Step 8: Cost Optimization Report

  COST OPTIMIZATION REPORT
  Current monthly spend:        $
  Projected after optimization: $
  Total monthly savings:        $ ()
  Annual impact:                $
  Savings breakdown:
  Waste elimination:     $ ( actions)
  Right-sizing:          $ ( resources)
  Pricing optimization:  $ ( reservations)
  Scheduling:            $ ( environments)
  Implementation effort:
  Quick wins ( savings
  Medium effort (1 week): $ savings
  Long-term (1 month+):  $ savings
  Risk: LOW — all changes are reversible

Step 9: Commit and Transition

1. Save report as docs/cost/-cost-optimization.md
2. Commit: "cost: — $/month identified ( recommendations)"
3. Provide actionable next steps with priority order

# Check cloud cost reports
curl -s http://localhost:8080/api/costs/summary | jq .total
grep -r "instance_type" infra/ | head -5

Key Behaviors

# Analyze cloud costs
aws ce get-cost-and-usage --time-period Start=2026-02-01,End=2026-03-01 --granularity MONTHLY --metrics BlendedCost
infracost diff --path .

1. Data-driven only. Actual utilization data, not assumptions.
2. Dollar impact required. "$180/mo savings" not "oversized".
3. Risk assessment. LOW/MEDIUM/HIGH per recommendation.
4. Reversibility matters. Right-sizing > reserved purchases.
5. Environment awareness. Conservative for prod, aggressive for dev.
6. Tagging is foundational. Fix tags before optimizing.
7. Continuous, not one-time. Alerts + monthly review.

On failure: revert with git reset --hard HEAD~1.

Flags & Options

| Flag | Description | |--|--| | (none) | Full cost analysis and optimization report | | --provider | Target specific cloud provider | | --scope | Narrow analysis scope | | --waste | Waste detection only | | --rightsize | Right-sizing recommendations only | | --tags | Cost allocation tagging audit only | | --budget | Budget alert configuration only | | --quick | Top 10 savings opportunities, skip deep analysis | | --report | Generate report from last analysis | | --threshold | Only show savings above threshold |

HARD RULES

1. NEVER STOP until all resource categories are analyzed and all savings are quantified in dollars.
2. EVERY recommendation MUST include dollar impact — "oversized" is not actionable, "$180/month savings" is.
3. EVERY recommendation MUST include risk level and reversibility assessment.
4. NEVER recommend reserved instances for workloads with less than 3 months of stable data.
5. NEVER apply dev-level aggressive optimization to production resources.
6. ALWAYS fix tagging first — cost optimization without attribution is guesswork.
7. git commit BEFORE verify — commit the cost report, then verify recommendations.
8. TSV logging — log every cost analysis:

`` timestamp provider scope current_spend projected_savings recommendations quick_wins ``

Explicit Loop Protocol

When analyzing resources across categories:

current_iteration = 0
resource_categories = [compute, storage, database, network, containers, serverless, other]
all_recommendations = []

WHILE resource_categories is not empty:
    current_iteration += 1
    category = resource_categories.pop(0)

    # Inventory
    resources = list_resources(category)

    FOR each resource in resources:
        utilization = get_utilization(resource, period="14d")

        IF utilization.cpu_avg /dev/null && echo "aws"
   gcloud config get-value project 2>/dev/null && echo "gcp"
   az account show 2>/dev/null && echo "azure"

2. Infrastructure as code:
   ls terraform/ *.tf 2>/dev/null && echo "terraform"
   ls pulumi/ Pulumi.yaml 2>/dev/null && echo "pulumi"
   ls cdk.json 2>/dev/null && echo "cdk"

3. Resource inventory tools:
   which aws-nuke cloud-nuke infracost 2>/dev/null

4. Existing cost tools:

Output Format

Print on completion: Cost: ${current_monthly}/mo → ${projected_monthly}/mo (-${savings}/mo, -{savings_pct}%). Top waste: {top_waste}. Untagged: {untagged_count} resources. Reservations: {ri_recommendation}. Verdict: {verdict}.

Keep/Discard

KEEP if: improvement verified. DISCARD if: regression or no change. Revert discards immediately.

Stop Conditions

Stop when: target reached, budget exhausted, or >5 consecutive discards.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: arbazkhan971
• Source: arbazkhan971/godmode
• License: MIT

Install and usage instructions live in the source repository linked above.