Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →dautovri/cloud-cost-agent
Use when analyzing cloud spend, running a cost audit, or finding savings across AWS, Google Cloud, and Azure — rightsizing, idle resources, and commitment/reservation recommendations, per-provider or cross-cloud. Uses only native CLIs and recommendation services (Cost Explorer/Compute Optimizer/Cost Optimization Hub, GCP Recommender, Azure Advisor + Cost Management) to produce prioritized savings reports.
Passed 21 of 22 test cases with the skill, and 18 of 22 without it.
and 2 more in the full report
3 fixed18 passed either way1 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 3 test cases the skill fixed. Same model, same prompt, with and without the skill.
We need to audit our primary AWS account for cost reduction opportunities over the past month. Please check native AWS recommendation engines to identify underutilized infrastructure and rightsizing potential. Deliver a structured summary outlining total potential savings grouped by implementation effort, plus the specific safe command to inspect the top priority item.
| Checked | Without skill | With skill |
|---|---|---|
| Uses the us-east-1 region specifically when invoking Cost Optimization Hub recommendations | Missed | Passed |
| Queries list-recommendations from cost-optimization-hub rather than relying on Trusted Advisor | Passed | Passed |
| Filters recommendations using implementation effort categories like VeryLow or Low | Passed | Passed |
The instructions your model reads when the skill loads.
Unified agent skill for native cloud cost optimization across the top 3 providers.
Requirements: AWS CLI v2, gcloud, and/or Azure CLI configured with read access to billing/recommender/advisor services. Several sources require one-time opt-in: AWS Compute Optimizer and Cost Optimization Hub (enrollment), GCP Recommender (API enabled, ~24–48h of data). jq is used in a few GCP filters.
Core rule: Always use the provider's official CLI and recommendation engines. No third-party tools.
aws ec2 terminate-instances/stop-instances, gcloud compute instances delete, az vm delete/deallocate, any purchase/create/modify) without quoting the exact command to the user and getting explicit approval.--profile, gcloud --project, az account set).--query / --format json / -o json for clean output.NextToken/nextPageToken, or use --no-paginate/--max-items (AWS) and --page-size/--limit (gcloud) deliberately. A single-page read gives wrong totals.aws ce (Cost Explorer) charges ~$0.01 per request. Do not loop Cost Explorer calls; cache results in the session and cap at a handful of calls per audit.BSD/macOS and GNU/Linux date differ. Compute the window once, portably, then reuse $START/$END:
START=$(date -u -d '30 days ago' +%Y-%m-%d 2>/dev/null || date -u -v-30d +%Y-%m-%d)
END=$(date -u +%Y-%m-%d)
(Or just have the agent insert literal YYYY-MM-DD dates.)
Tell the agent the provider(s) and goal. Examples:
audit AWS last 30 daysfind quick wins on GCP and Azurerightsizing recommendations for all cloudsAWS
aws sts get-caller-identity
aws ce get-cost-and-usage \
--time-period Start=$START,End=$END \
--granularity MONTHLY --metrics "UnblendedCost"
test cases passed, vs 18 / 22 without
Published skills whose descriptions closely match this one, with their headline benchmark lift.
More skills from the same source and nearby in the registry, with their headline benchmark lift.