How to use the Cloud Cost Optimization agent skill

What is an agent skill?

An agent skill is a folder containing a SKILL.md file with instructions and, optionally, reference files that an AI agent loads when a task matches the skill's description. The agent reads the description to decide when the skill is relevant, so you do not have to paste the instructions into every chat.

This skill's SKILL.md covers cost visibility, right-sizing, pricing models, architecture changes and governance, and it links to a tagging standards reference.

What the skill makes the agent do

Once installed, the agent can draw on a framework of four areas: visibility (cost allocation tags, budget alerts, dashboards), right-sizing (finding over-provisioned or idle resources), pricing models (reserved capacity, savings plans, spot or preemptible instances, committed use discounts) and architecture (managed services, lifecycle policies, caching).

It also brings provider-specific notes for AWS, Azure, GCP and OCI, sample Terraform for tagging, budgets, S3 lifecycle rules and auto-scaling, and a checklist of common savings actions.

• Suggest right-sizing of resources based on utilization
• Recommend cost-allocation tags, with a standard set of required tags
• Compare reserved, savings-plan, committed-use and spot/preemptible options
• Add lifecycle rules to storage
• Set budget alerts and review anomaly detection tools

Install the skill

Download the ZIP from the skill's page on this site and unzip it. For a personal install, place the resulting cost-optimization folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

If you use the Claude apps instead, open Settings, go to Capabilities, and upload the skill under Skills.

• Personal: ~/.claude/skills/cost-optimization/
• Project: .claude/skills/cost-optimization/ inside your repository

Trigger the skill with natural requests

You do not need a special command. Describe the task in plain language that matches the skill's purpose: reducing cloud expenses, analyzing infrastructure costs, or implementing cost governance policies. Naming the provider and the artifact you want helps the agent pick the right section.

Good requests name a provider, a goal and an output format, for example asking for a Terraform budget with an 80 percent alert, or a list of rightsizing candidates from a description of your instances.

• "Review our AWS setup for cost savings and list the levers in priority order."
• "Write Terraform that tags every resource with Environment, Owner, CostCenter, Project and ManagedBy."
• "Add an S3 lifecycle rule that moves objects to infrequent access after 30 days."
• "Draft a monthly budget with an alert at 80 percent of actual spend."
• "What reserved instance or savings plan options fit a steady production workload on Azure?"

Example workflows

A typical session starts with visibility. Ask the agent for a tagging scheme, then for the Terraform locals block that applies it. Next, ask which pricing model suits steady versus bursty workloads, and where spot or preemptible capacity could fit with a fallback on standard capacity.

For storage, the skill includes a three-tier example (Standard, Standard-IA at 30 days, Glacier at 90 days, Deep Archive at 365 days). Treat these as ideas to adapt, not defaults for your data.

Check the output before you use it

The skill is guidance and sample code. It does not connect to your cloud account, read billing data or change anything by itself. The Terraform examples use placeholders: a hard-coded AMI ID (ami-12345678), a budget start date of 2024-01-01 and example email addresses. Replace them and review every change before applying it.

The discount percentages in the skill (for example up to 72 percent for reserved VMs or up to 90 percent for spot) are generic figures from vendor marketing. They may be outdated or differ from your contract, region or usage, so do not use them as a savings forecast. Automated resizing or shutdown can cause outages, so review changes in a non-production environment first.

• Terraform is needed only if you ask the agent to write or apply the infrastructure examples
• Provider accounts and credentials are needed only if you later ask the agent to query real billing or resource data; the skill itself does not require them
• The skill runs no scripts and uses no network access

Tips for better results

Give the agent real context: your provider, the services that dominate your bill, and whether workloads are steady or bursty. Generic questions get generic checklists.

Keep the reference file in mind. The tagging standards file defines required tags (Environment, Owner, CostCenter, Project, ManagedBy) and recommends an approved tag dictionary, policy enforcement and weekly audits. Asking the agent to follow it keeps outputs consistent across teams.

• Name the provider and the service (for example EC2, RDS or Compute Engine)
• Ask for a prioritized list so you know where to start
• Ask the agent to flag placeholder values in any code it writes