Set up
Connect your tools
Claude is more powerful when it works directly with your systems. You control permissions and access. Learn about tool access(opens in new tab).
Add a connector(opens in new tab) for each tool you want Claude to use.
Set your working folder
Drop the billing CSV, your tag-to-team mapping, and last month's report into one folder, then give Claude access to it(opens in new tab) in the Claude desktop app. With these files, Claude has the baseline and the ownership map. The anomaly report gets written back here. Create a project(opens in new tab) from the folder so your thresholds and known-okay exceptions persist month to month.
The prompt
Copy this into Claude
Compare this month's AWS costs to the trailing three-month average by service and tag:team. Flag any line more than 20% over trend, trace each to the workload or change behind it, and write the cost-anomaly report with an owner and recommended action for every item. Put a one-line summary for the infra channel at the top.
Why this works
Set the comparison baseline. One month is noise; three months is a baseline.
Set the threshold. "More than 20% over" turns a wall of numbers into a short list worth reading.
Ask for what caused each anomaly. "Likely cause and owning team" is what makes a finding useful — so each anomaly includes where it came from and who to talk to about it.
Include who owns what. Every anomaly has an owner, so the report is a to-do list, not an FYI.
Get a better draft
Add an example to match. Drop an example you like into the folder and Claude matches your structure and voice.
Ask it to flag uncertainty. Add "flag anything you're not confident about" so you know where to look first when you review the draft.
Finish it where the file lives
Make Claude work for you
Tell Claude your own rules for this task, such as your tag scheme and the spikes you've already accepted, then save what you've done as a skill of your own. A few minutes of conversation and it runs with your standards from then on.
Make what we've done in this task so far into a skill called /cost-anomalies.
Review the skill, and save it if Claude hasn't already.
Make it repeatable
Run it when the bill closes
The report should be waiting before anyone opens the console. Ask Claude to set up a scheduled task, or open Scheduled in the sidebar, and /cost-anomalies runs against the new export on the first Monday after it's in the folder.
Set up a scheduled task for every Monday at 7am: if a new month's billing export is in Infra/cloud-cost, run /cost-anomalies against it, write the anomaly report to that month's folder, and put the one-line summary for the infra channel at the top.
Runs /cost-anomalies on the closed month, writes the ranked anomaly report to the dated folder, and puts the one-line summary for the infra channel at the top.
To use a folder on your computer during a scheduled task, keep your computer on and awake with the Claude desktop app open so Claude can access it.
Share with your teammates
Your /cost-anomalies skill now knows your tag scheme, your threshold, and the spikes you've already accepted. Share it so any infra engineer can run the same analysis mid-month when something looks off.