Skills Reference

Skills are guided analysis workflows available through the CloudZero Plugin for Claude Code. Each skill combines multiple MCP server tools into a structured investigation. Skills are automatically invoked by Claude based on your prompt. You can also invoke them directly using the /skill-name slash command syntax in Claude Code.

All skills depend on organization context. The understand-cloudzero-organization skill runs first and caches the context for all subsequent skills. The diff-cost-projection and cost-projection skills also require a Git repository with infrastructure-as-code files.

Allocation

tag-coverage-analysis

Evaluates cloud resource tagging quality and coverage to identify untagged or poorly tagged resources and improve cost allocation accuracy.

When to use

• Setting up showback or chargeback
• Governance and compliance reviews
• Identifying unattributed costs
• Measuring tagging policy effectiveness

Trigger keywords

tag, tagging, untagged, coverage, attribution, showback, chargeback, governance

Example prompts

What's my tag coverage?

Show me untagged costs by account

Analyze tagging quality for showback accuracy

Output includes

• Overall coverage percentage and tag scorecard
• Untagged cost breakdown by service, account, and region
• Tag value distribution and quality issues (inconsistent values, typos)
• Coverage trends over time
• Prioritized tagging recommendations

Critical tags typically analyzed

• Environment (production, staging, development)
• Team or Owner
• Application
• CostCenter
• Project

Analytics

cost-comparison

Compares cloud costs across different time periods, environments, accounts, regions, teams, or other Dimensions to understand variations and benchmark efficiency.

When to use

• Comparing periods (this month vs. last month)
• Benchmarking environments (production vs. staging)
• Evaluating team efficiency
• Measuring before/after optimization impact

Trigger keywords

compare, comparison, versus, vs, difference, between, benchmark, relative

Example prompts

Compare this month's costs to last month

How do production costs compare to staging and development?

Compare Team A's cloud spending to Team B

Output includes

• Total costs for each group with key differences
• Service-by-service comparison
• Items unique to each group
• Normalized per-unit metrics when scale differs
• Recommendations for improving higher-cost groups

cost-trend-analysis

Analyzes cloud cost trends over time to identify spending patterns, growth rates, and seasonality for budget planning and forecasting.

When to use

• Understanding cost growth trajectory
• Budget planning and forecasting
• Quarterly or annual reviews
• Predicting future spending

Trigger keywords

trend, growth, trajectory, pattern, forecast, over time, month over month, increasing, decreasing, velocity

Example prompts

How are my costs trending over the last 90 days?

Forecast next month's spending based on current trends

What's my month-over-month cost growth rate?

Output includes

• Trend direction (growing, stable, or declining)
• Growth rates (week-over-week, month-over-month, quarter-over-quarter)
• Pattern analysis (linear, exponential, seasonal)
• Dimensional breakdown of which services or accounts drive the trend
• Projected future costs with confidence ranges

Granularity guidelines

• 1-3 months: Use daily granularity
• 3-6 months: Use weekly granularity
• 6+ months: Use monthly granularity

custom-dimension-analysis

Analyzes costs using organization-specific custom Dimensions (such as teams, products, features, or business units) for business-aligned cost visibility and showback/chargeback reporting.

When to use

• Showback and chargeback reporting
• Product P&L analysis
• Team budget tracking
• Feature cost attribution

Trigger keywords

team, product, feature, business unit, application, custom, showback, chargeback

Example prompts

Show me costs by team for the last month

What's each product's cloud spending?

Generate a showback report by team

Output includes

• Cost distribution by Dimension values
• Growth trends for each value
• Service mix per team or product
• Unallocated costs that cannot be attributed
• Showback report formatted for finance and business teams

service-cost-deep-dive

Performs detailed analysis of a specific cloud service, breaking down by usage types, resources, regions, and accounts while identifying service-specific optimization opportunities.

When to use

• Targeted optimization for a specific service
• Understanding service cost composition
• Finding rightsizing opportunities
• Analyzing Reserved Instance or Savings Plan coverage

Trigger keywords

deep dive, analyze, breakdown, detailed, specific service, EC2, RDS, S3, Lambda

Example prompts

Do a deep dive on my EC2 costs

Analyze my RDS spending in detail

Show me optimization opportunities for my compute services

Output includes

• Service overview with total cost, trend, and share of total spend
• Distribution by region, account, and usage type
• Time-based usage patterns
• Service-specific optimization recommendations
• Current savings and additional savings potential

top-cost-drivers

Identifies and ranks the highest cloud cost contributors across services, accounts, teams, regions, or custom Dimensions to help prioritize optimization efforts.

When to use

• Monthly cost reviews
• Budget planning and allocation
• Identifying optimization priorities
• Answering "where is my money going?"

Trigger keywords

top, biggest, largest, highest, most expensive, cost drivers, where is money going, prioritize

Example prompts

What are my top 10 cost drivers this month?

Show me the biggest costs by service

Rank my cloud spending by team

Output includes

• Total spend with top contributors ranked
• Percentage contribution and cumulative analysis (80/20 identification)
• Multi-dimensional breakdown (services within accounts)
• Optimization priorities based on impact

Anomalies

cost-anomaly-detection

Scans for cost anomalies, unusual spending patterns, and unexpected changes that indicate waste, misconfiguration, or optimization opportunities.

When to use

• Proactive cost monitoring
• Weekly or monthly cost reviews
• Waste identification
• Before presenting cost reports

Trigger keywords

anomaly, unusual, abnormal, irregular, unexpected, odd, suspicious, detect issues

Example prompts

Are there any cost anomalies?

Scan for unusual spending patterns

Check for abnormal costs this week

Output includes

• Anomaly count by severity (high, medium, low)
• Detailed analysis of each anomaly with impact and cause
• Time-series visualization showing when anomalies occurred
• Newly created high-cost resources
• Security concerns (patterns that may indicate issues)
• Prioritized remediation steps

Anomaly types detected

• Spikes: Sudden increases that return to normal
• Step changes: Sudden persistent increases
• Gradual drift: Slow increases over time
• New appearances: Resources that did not exist before
• Disappearances: Resources that suddenly stopped
• Pattern breaks: Usage patterns that changed

cost-spike-investigation

Investigates sudden cost increases by comparing recent spending to baseline periods and identifying which services, accounts, or resources are responsible for the change.

When to use

• Unexpected cost increases reported
• Monthly bills show significant jumps
• Alerts indicate spending anomalies
• Investigating budget overruns

Trigger keywords

spike, increase, jump, surge, anomaly, unexpected, sudden change, why did costs go up

Example prompts

Why did my costs increase by 30% this month?

Something caused a cost jump on Tuesday. Find out what it was.

Investigate the cost anomaly in my production account

Output includes

• Total increase and primary cause
• Baseline vs. spike period comparison
• Root cause analysis identifying services and accounts driving the increase
• Time-series analysis showing when the spike started and whether it is ongoing
• Recommendations to address or prevent recurrence

Foundation

understand-cloudzero-organization

Retrieves organization-specific context from CloudZero that all other skills depend on. This skill runs automatically at the start of any analysis session and caches the results for subsequent skills.

When to use

• At the start of any cost analysis conversation (runs automatically)
• When switching to analyze a different organization
• When you need to understand available custom Dimensions

Example prompts

Load my CloudZero organization context

What custom dimensions are available in my organization?

Output includes

• Custom Dimensions with business context for each
• Organization-specific analysis workflows and naming conventions
• Cost allocation policies for shared cost distribution

Key behaviors

• Caches results within a conversation and does not re-fetch
• All other skills automatically reference the cached context
• Runs once per session unless switching organizations

Optimization

cost-projection

Reads infrastructure-as-code definitions and projects what they will cost to run. Enumerates every resource, maps each to CloudZero Dimensions, queries actual spend for existing resources, estimates costs for new ones, and produces a line-item breakdown with monthly totals.

When to use

• Before deploying new infrastructure
• Evaluating architecture options by cost
• During design reviews to compare cost of alternatives
• Estimating the monthly cost of a Terraform module, CDK stack, or CloudFormation template

Trigger keywords

project cost, how much will this cost, estimate deployment, monthly cost, infrastructure cost, stack cost

Invocation

/cost-projection [target]

Where [target] is one of:

• A file path: main.tf, lib/my-stack.ts, template.yaml
• A directory: terraform/, cdk/lib/ (analyzes all IaC files within)
• Omitted: auto-discovers all infrastructure in the codebase

Supported frameworks

• Terraform / OpenTofu
• AWS CDK (TypeScript, Python)
• CloudFormation
• SAM
• Serverless Framework
• Pulumi

Example prompts

How much will this stack cost?

What's the monthly cost of this Terraform module?

/cost-projection terraform/production/

Output includes

• Cost summary grouped by category (compute, database, storage, messaging, networking) with monthly estimates
• Per-resource line items with type, configuration, monthly estimate, and data source
• Assumptions and caveats (region, pricing model, unresolved variables)
• Optimization opportunities (RI/SP recommendations, rightsizing, architecture alternatives)

Confidence levels

• HIGH: Fixed-cost resources with specific configuration and pricing verified via CloudZero data or pricing lookup
• MEDIUM: Fixed-cost resources using fallback formulas, or resources with some usage-dependent component
• LOW: Fully usage-dependent resources (such as S3, DynamoDB on-demand, data transfer) where cost scales with unknown traffic

diff-cost-projection

Analyzes code diffs for infrastructure cost impact. Detects changes to Terraform, CDK, CloudFormation, SAM, Kubernetes, scaling configurations, and application code that affect cloud spending. Queries CloudZero for current spend baselines on affected services and produces cost impact estimates with confidence levels.

When to use

• Before merging infrastructure or application code changes
• During code review to flag cost implications
• Estimating the cost impact of a PR or branch

Trigger keywords

cost impact, diff, PR cost, branch cost, infrastructure changes, will this increase costs

Invocation

/diff-cost-projection [target]

Where [target] is one of:

• Omitted: analyze all changes on the current branch vs. its base (default)
• feature/my-branch: analyze branch diff against base
• abc123..def456: analyze a commit range

Example prompts

What's the cost impact of my changes?

Will this branch increase our cloud spend?

/diff-cost-projection feature/new-cluster

Output includes

• Verdict: cost increase, decrease, mixed impact, or no impact
• Summary table with one row per change showing service, current spend, estimated impact, and confidence
• Net estimated impact with monthly and annualized figures
• Per-change breakdown with file and line references, before/after values, and reasoning

Confidence levels

• HIGH: Direct resource change with clear before/after values and solid baseline data
• MEDIUM: Scaling changes with predictable multipliers, or direct changes without baseline
• LOW: Indirect application changes, complex interactions, or estimates dependent on runtime behavior

optimize-triage

Fetches top unaddressed recommendations from CloudZero Optimize, researches each one for context, and surfaces only the recommendations that are genuinely actionable. This skill is read-only and does not make changes to resources or infrastructure.

When to use

• During weekly or monthly cost reviews
• Prioritizing which recommendations to act on first
• Getting deeper context on CloudZero Optimize recommendations

Trigger keywords

optimize, recommendations, savings, cost optimization, waste, triage, what should I optimize

Example prompts

What are the top optimization opportunities?

Triage my CloudZero Optimize recommendations

Find the biggest cost savings I should act on

Output includes

• Qualifying recommendations ranked by savings with resource, account, impact, effort level, and confidence verdict
• Final summary with claimed vs. realistic savings, confidence verdicts, effort, suggested contacts, and next steps
• Per-recommendation report files written to ./optimize-triage-reports/

Research performed per recommendation

Each recommendation is investigated in parallel:

• IaC search: Checks if the resource is managed by Terraform, CDK, or other IaC
• Git history: Searches for recent changes, related PRs, and commit authors
• Signal sweep: Searches connected tools (Jira, Slack) and CloudZero cost trends for context
• Cloud provider state: Queries current resource configuration and utilization (read-only)
• Category-specific checks: Tailored analysis based on recommendation type (idle resources, rightsizing, reservations, storage tiering)

Confidence verdicts

• HIGH: Multiple corroborating signals, no contradictions, clear owner
• MEDIUM: Mostly corroborated with one or more open questions
• LOW: Significant contradictions or missing context (suppressed from output)
• SKIP: Clear "do not act" signal found (excluded from output)