Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Data Analysis Mode

    1

    A reproducible five-stage data-analysis workflow, with five hands-on references and a working Excel formula-recalculation script.

    Free

    6 usesSecurity scanned
    Data Analysis Mode

    Data Analysis Mode

    Example session with this skill installed

    Analyse monthly_sales.csv: compare 2024 revenue across the three regions, chart it, and tell me whether the gap is about scale or about pricing.

    • Read your context and instructions
    • Compiled the data analysis mode
    • Generated the data export

    Data quality first — 37 raw rows → 36 clean

    Three problems found, all handled inside a reproducible script:

    IssueWhereHandling
    Missing revenueEast, 2024-03Linear interpolation → 129,500 (neighbours 117,800 → 141,200)
    Duplicate rowNorth, 2024-07Deduplicated, one record kept
    Missing metricAll rowsDerived avg order value = revenue ÷ orders

    Cross-check: the interpolated East March figure implies an order value of 159.5, almost identical to East's annual median of 159.4 and consistent with the 812 orders recorded that month.

    Region comparison, 2024

    RegionRevenue (CNY)ShareAvg order valueCVH2 vs H1
    East1,835,20043.1%159.20.173+27.9%
    South1,405,60033.0%160.10.169+26.8%
    North1,022,10024.0%152.00.130+20.0%

    What the numbers say

    • The gap is structural, not a pacing problem: the three regions correlate at ≥0.996 month over month, so the ranking never shifts on its own.
    • East leads North by 79.6%. East and South have almost identical order values (159.2 vs 160.1), so that 30.6% gap is volume, not pricing — 11,527 vs 8,779 orders.
    • North's problem is price and momentum, not just volume: lowest order value (152.0, 5.1% below South) and the weakest second-half growth (+20.0%).

    Deliverables written to the workspace

    analyze_sales.py (reproducible pipeline with cleaning comments), region_revenue_comparison.png (200 dpi, four panels), region_summary.csv, analysis_report.md.

    Limitations

    The missing March figure is an estimate and is flagged in the chart. One year of monthly data cannot separate seasonality from trend — a second year is needed to test whether the North slowdown is cyclical.

    Next steps

    1. Pull the same breakdown for 2023 to test whether North's H2 deceleration is seasonal or persistent.
    2. Split North's orders by product tier to confirm the low average order value is a mix problem.
    3. Choose the intervention from that mix analysis rather than from the headline gap.

    region_revenue_comparison.png

    CSV · data export

    Generated

    Example file from a real run - the skill writes it into your workspace.

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

    About this skill

    A reproducible five-stage workflow for real datasets — scope the question, profile and clean the data, run EDA, model it, then write a conclusion you can defend — plus five hands-on references that do the actual work: pandas 3.x behaviour changes and patterns, Excel workbook I/O with a working formula-recalculation script, safe MySQL/PostgreSQL access through SQLAlchemy, hypothesis testing and scikit-learn modelling with assumption checks, and chart standards for matplotlib/seaborn. Includes SKILL.md (the workflow plus a delivery checklist), five reference files under references/, a working scripts/recalc_excel.py, and a README.

    • Requires Python 3.10+ with pandas and openpyxl. The Excel formula-recalculation script needs Microsoft Excel or LibreOffice installed, and the Excel COM path is Windows-only.
    • The database reference covers MySQL and PostgreSQL only, and you supply your own read-only credentials.
    • It structures and audits the analysis; it does not replace domain judgement — conclusions still need your review.

    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

    6 uses

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    • Free with an Agensi account

    Requires

    Mysql
    Postgresql
    Python

    Permissions

    Terminal / Shell
    Environment Variables
    Read Files
    Network Access
    Write Files

    Allowed Hosts

    mysql.com
    postgresql.org

    File Scopes

    **/*.csv
    **/*.xlsx
    **/*.xlsm
    **/*.parquet

    Runs local Python. The Excel formula-recalculation script uses Excel COM when available and falls back to LibreOffice.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    It makes your agent run data analysis in a fixed, checkable order: confirm the question and metric definitions, profile and clean the data in a script, run EDA, check modelling assumptions, then write a conclusion with stated limitations. Five reference files cover the tooling, and a delivery checklist keeps the result traceable.

    Eight files: SKILL.md (five-stage workflow plus a delivery checklist), five references (pandas-3x.md, excel-workbooks.md, database-access.md, statistical-modeling.md, visualization.md), scripts/recalc_excel.py, and a README.md. MIT licensed.

    No. Only the formula-recalculation script needs a spreadsheet engine: it uses Excel via COM when available and falls back to LibreOffice. That COM path is Windows-only; the rest of the skill is plain Python and platform-independent.

    Yes, for MySQL and PostgreSQL, through SQLAlchemy with read-only credentials you supply via environment variables — the reference keeps passwords out of code and out of logs. It checks table structure and applies LIMITs before pulling large tables, and it does not write to your database.

    No hard limit is built in, but pandas is memory-bound: for very large files the skill will suggest DuckDB, polars or chunked reads rather than pretending it can load everything. For database work it starts with a LIMIT and selects only the columns needed.

    It removes the improvisation: stages run in order, cleaning steps are written into a script rather than applied by hand, statistical assumptions are checked before conclusions, and every analysis ships with a data-quality audit and its limitations. That is what makes the result reproducible instead of a one-off answer.

    Any agent that loads SKILL.md folders — Claude Code, Codex CLI, Cursor, GitHub Copilot, Gemini CLI and others. No vendor-specific integration and no account to connect.

    There is no auto-update mechanism: the reference documents pandas 3.x behaviour explicitly, so a future version that changes those semantics could make it out of date until revised. It is MIT licensed, so you can fork and adjust.

    New in Data & Databases