Cloud Bill Analyst: AWS, GCP and Azure cost review

    1

    Find out why your cloud bill moved, from your own CUR, GCP or Azure export, fully offline.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Cloud Bill Analyst: AWS, GCP and Azure cost review

    Example session with this skill installed

    Our AWS bill more than doubled in October and finance wants to know why. I've downloaded the cost and usage reports for September and October. We tag resources by team.

    • Read your context and instructions
    • Compiled the cloud bill analyst

    Plan: check both reports with --inspect, run the review on the unblended basis (it matches the invoice) with --tag-key team, then rerun amortized to separate commitment purchases from usage.

    1. The headline (USD, unblended)

    MonthChargesCreditsTaxNet
    2026-091,853.29-250.0095.401,698.69
    2026-104,228.05-250.00241.184,219.23

    Both months complete, one currency, 0 skipped rows. Charges rose 2,374.76:

    ServiceSepOctChange
    ComputeSavingsPlans0.001,800.00+1,800.00
    AmazonAthena90.00500.00+410.00 (+455.6%)
    AmazonEC21,145.291,289.45+144.16 (+12.6%)

    2. Spikes, and a rerun

    The script flagged 2026-10-01 at 29.99x the trailing median, driven by ComputeSavingsPlans 1,800.00: a Savings Plan paid upfront that day. A flag on the 1st is usually a fee, so I reran with --cost-basis amortized:

    2026-10 charges          4228.05 -> 2580.88
    month-over-month change  2374.76 -> 727.59
    2026-10-01 no longer flagged; amortized_fallback_rows: 0
    

    Two real changes remain in both views

    2026-10-14: 459.56 against a 52.56 median (8.74x). AmazonAthena cost 410.00 against a usual 3.00. One service, one day: look for a runaway query.

    2026-10-20 to 10-25: flagged daily at 1.62x, AmazonEC2 +32.64 a day. That is one step up (a new c6i.2xlarge workload tagged search), not six incidents.

    3. Allocation and commitments

    • Untagged: 3,595.16 of charges (59.1%). Half of that is the Savings Plan fee (1,800.00), which can't carry a tag; amortized, untagged is 40.5%.
    • Savings Plan coverage in October: 27.4% of eligible usage at on-demand-equivalent cost (285.70 covered, 757.42 not).

    4. Likely waste (each one: check before acting)

    LeadEvidenceCost, 2 months
    Idle NAT gateway nat-0dd44ee55ff6600021,464.0 billed hours, 3.050 GB processed66.02
    Older m4 instance family1 instance, 61 rows292.80
    EBS gp2 volumes3 volumes305.00
    EBS snapshot1 snapshot152.50

    The bill can't show whether a volume is attached or the NAT is a standby.

    Next steps

    • Check the Athena query history for 14 October before it repeats.
    • Confirm the search workload from 20 October was planned; it is the lasting part of the increase.
    • Want October alone on the amortized basis, or untagged spend by account? Ask, and I'll rerun with those options.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Drop in an AWS Cost and Usage Report (legacy or CUR 2.0, CSV or gzip), a Cost Explorer download, a GCP billing export, an Azure cost details file or any date/service/cost CSV. A local standard-library Python script computes every number with exact decimal maths; no cloud credentials, no API calls. You get spend by month, top services, accounts, regions and usage types, the biggest month-over-month movers, spike days with the services behind them, untagged spend for your tag key, credits and tax kept separate, and Savings Plan, RI and CUD coverage. Possible waste (storage volumes and snapshots, NAT gateways and load balancers moving little data, older instance families) is offered as leads to check, with evidence, never as findings.

    Works only on billing exports you already have; no API access. Waste leads are hypotheses to check. Commitment coverage needs the right columns and is not computed for Azure. Very large reports are read in one pass, so filter with --from/--to or pass one month at a time. Overlapping exports are not de-duplicated.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 9 days ago

    • Free with an Agensi account

    Requires

    Python 3

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    Runs the bundled scripts locally on the files the user names. Writes output only inside the user's project, after asking, and makes no network requests.

    Listed9 days ago

    What's inside

    Frequently Asked Questions

    The skill processes local CSV and .gz exports from AWS (CUR/CUR 2.0), GCP, and Azure. This offline approach allows you to analyze massive billing files that exceed standard LLM context windows while ensuring sensitive cost data stays within your local environment.

    A Python script does the arithmetic with exact decimal maths over every row, and the agent only explains the script's output and cites the bundled reference. The script ships with a self-test covering the export formats, gzip input, month-over-month movers, planted spike days, untagged spend, credits and currency handling.

    No, this skill is designed for privacy and security; it does not require API keys or credentials for your cloud consoles. You simply provide the exported billing files (CSV or .gz) from your cloud provider, and the skill performs all analysis locally.

    No. It never calls cloud APIs and cannot change anything. It lists possible waste as leads to check, with the evidence behind each: storage volumes and snapshots, NAT gateways and load balancers moving little data, and older instance families. A bill cannot show whether a volume is attached or a gateway is a standby, so verify each one in your console before acting.

    SKILL.md (the agent's workflow and report format), scripts/analyze.py (the analyzer, Python standard library only) and reference.md (what each provider's columns mean, amortized vs unblended cost, and how to read spike days and coverage).

    New in DevOps & Cloud