Cost Optimization (Cloud FinOps)

Provides a deterministic workflow to reduce cloud spend safely while maintaining reliability and performance.

Use this skill when

• A team needs to reduce cloud spend (quick wins + longer-term program)
• A team needs tagging/label standards and cost allocation
• A team needs budgets, alerts, and anomaly detection
• A team needs rightsizing and commitment strategy (RIs/Savings Plans/CUDs)
• A team wants unit economics (cost per request/job/tenant)

Do not use this skill when

• The request is for cloud architecture/platform selection or migrations
• The request is for CI/CD or deployment mechanics
• The request is for database-specific tuning and cost/performance tradeoffs

Inputs to collect

• Billing scope (accounts/subscriptions/projects, envs, owners)
• Time window (last 7/30/90 days)
• Cost targets (budget, % reduction, unit-cost goal)
• Constraints (SLOs, compliance, procurement limits)
• Current tagging/labeling coverage and budget/alert setup

Workflow (Deterministic)

1. Confirm scope, owners, and constraints.
   • Output: scoped summary with success criteria and missing inputs.
   • Decision: if scope/constraints are unclear, ask for the missing details before continuing.
2. Establish cost allocation baseline (tags/labels, budgets, anomaly alerts).
   • Output: required tags/labels list + current coverage gaps.
   • Decision: if tag coverage is too low to attribute costs, define the tagging remediation plan before optimization.
3. Identify top cost drivers by category and owner.
   • Output: top drivers table with root causes and affected services.
   • Decision: if billing data is unavailable, request a cost export or dashboard access before continuing.
4. Build quick wins backlog (idle cleanup, egregious sizing, retention policies).
   • Output: prioritized quick wins with estimated savings ranges and risk level.
5. Build structural wins plan (autoscaling, commitment strategy, data flow changes).
   • Decision: only recommend commitments after stable utilization evidence.
   • Output: structural initiatives with prerequisites and timelines.
6. Define verification gates (performance, reliability, cost attribution).
   • Output: verification checklist and rollback steps.
7. Compile the final savings plan and reporting format.
   • Output: consolidated plan with owners, timelines, and verification gates.

Common pitfalls

• Recommending commitments before usage stabilizes.
• Ignoring tag/label gaps that prevent attribution.
• Chasing small wins while top drivers stay untouched.
• Cutting observability without measuring impact on incident response.
• Skipping verification gates and rollback plans.

Output Contract (Always)

• Top cost drivers with owners and root causes
• A prioritized plan (quick wins + structural wins) with estimated savings ranges
• Risks and verification gates (performance/reliability)
• Tagging/governance recommendations

Reporting format

• Summary (scope, targets, constraints)
• Top cost drivers (category, owner, root cause)
• Savings plan (quick wins + structural wins with estimates)
• Risks + verification gates (SLOs, rollback)
• Tagging/governance actions

Examples

Example request

"We need to cut our AWS bill by 20% in 60 days without hurting reliability. We have mixed prod and non-prod workloads and incomplete tagging."

Example response (excerpt)

• Scope: 3 accounts (prod/staging/dev), 60-day window, target 20% reduction
• Top drivers: EC2 on-demand in prod, CloudWatch logs, data egress from analytics
• Quick wins: shutdown non-prod nights/weekends, log retention to 14 days
• Structural wins: autoscaling for batch workers, Savings Plan after 30 days stable usage
• Verification: latency/error budgets, rollback plan for scaling changes

Resources (Optional)

• References index: references/README.md
• End-to-end playbook: resources/implementation-playbook.md