Data Report Dashboard

What you're buying: a download package (skill instructions + Python scripts) that runs on your own computer inside your own AI agent — it is not a cloud service.
You need: an AI agent that supports skills (e.g. Claude Code or Codex) — its subscription or API costs are yours; Python 3.9+ with pandas, openpyxl and matplotlib.
Third-party services: none. Everything runs offline; Chart.js (MIT) is bundled inside the package. Your data's column names and summary statistics are seen by your own AI agent.

One Excel/CSV file plus one question in — an interactive dashboard your boss can read out.

What you get

• 📊 dashboard.html — a single file that opens offline and can be emailed as-is
  • 3–6 KPI cards and 4–8 interactive charts (hover for values; every chart can expand into a value table)
  • A sortable, searchable data table
  • Light/dark toggle and mobile-friendly layout
• 🖼️ charts/*.png — one PNG per chart, ready for slides
• 📝 Key findings — 4–7 points plus suggested next steps, every number machine-checked
• 🧾 results.json — every computed result, fully traceable

Example

Survey (English): 420 responses → NPS −24.5, average satisfaction 3.58, satisfaction by plan, multi-select feature usage.

Sales: 2,000 e-commerce orders (Chinese column names). Question: which channel and category bring the most revenue, and how do monthly trends and returns look? Count only completed orders.

• 5 KPIs (revenue NT$1,932,323, 1,784 completed orders, AOV NT$1,083, return rate 6.6%, 3,230 units), 7 charts and a 500-row table — one 254 KB HTML file
• Findings such as "Shopee is 43.25% of revenue", "October is the peak" and "December data ends on 12-25, so the month is incomplete" — every number matched results.json
• All figures cross-checked with an independent pandas script: identical

Every chart is also exported as a PNG, ready for slides:

Try it — copy a prompt

• "Here's sales_2025.xlsx — which channels and products drive revenue, and what's the monthly trend? Completed orders only."
• "Build a dashboard from survey.csv: NPS, satisfaction by plan, and which features people use."

Why buy it

One payment, a dashboard for every report — no BI subscription, no per-seat fees, and your data never leaves your computer.

Who it's for

Managers, analysts, marketers, e-commerce sellers, researchers and anyone who gets "can you make this into a report?" every week.

Why you can trust the numbers

• The AI doesn't do the maths: sums, averages, shares and growth rates are all computed by pandas
• Findings are auto-checked: every number in the summary must exist in the computed results — otherwise output is refused and the bad numbers are listed
• Common traps caught automatically: an incomplete last month (not mistaken for a drop), ID columns that must not be summed, outliers, "NA" misread as missing, and more

Supported analysis

• Derived columns (e.g. revenue = price × quantity × (1 − discount)) and filters (e.g. completed orders only)
• Time trends (day, week, month, quarter, year; missing periods filled with 0), rankings, composition, 100% stacked, distributions, correlation
• Survey multi-select answers ("Reports;API" is split automatically, with % of respondents per option)
• Ratio KPIs such as NPS, return rate and conversion rate
• 4 colour themes; interface in English or Traditional Chinese, findings in any language

Requirements

• Python 3.9+
• pip install pandas openpyxl matplotlib
• Reads CSV, TSV and XLSX (encoding detected automatically, including UTF-8, Big5 and CP950)

Limitations (please read)

• This is descriptive analysis: no forecasting, significance testing or causal claims — the summary avoids "A caused B".
• Not financial, tax or investment advice.
• One file and one sheet at a time; no database connections.
• Findings are written by AI; the numbers are checked, but the interpretation is yours to judge.
• The data table embeds up to 500 rows (adjustable); very large data makes the HTML bigger.

Files

• SKILL.md — agent workflow
• scripts/profile_data.py — data health check (types, missing values, outliers, date range)
• scripts/build_dashboard.py — calculations, dashboard, PNGs and results.json
• scripts/check_numbers.py — summary number check
• scripts/vendor/ — Chart.js 4.4.1 (MIT)
• references/ — chart-choice guide, plan format, English and Chinese examples with sample data