| name | csv-to-report-generator |
| description | Transforms CSV data into structured, readable reports with summaries, key metrics,
tables, and narrative insights. Use this skill whenever a user says "turn this CSV
into a report", "generate a report from this data", "summarize this spreadsheet",
"create a data report from these numbers", "write a business report from this CSV",
"analyze this data and produce a report", or "convert this data into something
readable". Also activate when a user shares tabular data and wants it turned into
a polished document. Do NOT use for building interactive dashboards (use
dashboard-design-skill) or writing full data cleaning pipelines.
|
CSV to Report Generator
Turn raw CSV data into a clear, actionable report — with summary statistics, key
findings, formatted tables, and plain-language narrative.
When to Use
- You have tabular data and need a readable document for stakeholders
- Generating a one-time report from an export (Stripe, Google Analytics, Airtable, etc.)
- Turning a data dump into a weekly/monthly report
- Creating a summary for an executive who won't look at raw data
When NOT to Use
- Building interactive dashboards with charts (use
dashboard-design-skill)
- Ongoing automated reporting pipelines (different engineering task)
- Complex statistical analysis (use a data science tool)
Workflow
Step 1 — Receive the Data
Accept:
- Pasted CSV text (first 20 rows is usually enough for structure analysis)
- Column names and a description of the dataset
- A file path (if file access is available)
Ask:
- What does this data represent?
- Who is the audience? (executive, analyst, customer, internal team)
- What's the reporting period? (daily, weekly, monthly, ad hoc)
- Are there specific metrics or questions to highlight?
Step 2 — Profile the Data
For each column, identify:
- Data type (numeric, categorical, date, text)
- Key statistics for numeric columns: min, max, mean, median, total
- Value distribution for categorical columns: top 5 values and counts
- Completeness: % non-null
Step 3 — Generate Insights
Look for:
- Top performers: Highest values, most frequent categories
- Outliers: Values significantly above or below average
- Trends: If date column exists, directional change over time
- Comparisons: Group-by summaries (e.g., by region, product, or team)
Step 4 — Structure the Report
# [Report Title]
**Period:** [Date range] | **Generated:** [Date] | **Source:** [Dataset name]
---
## Executive Summary
[3–5 sentence narrative summary of the most important findings]
## Key Metrics
| Metric | Value | vs. Previous Period |
| ------------------- | ------ | ------------------- |
| Total Revenue | $X,XXX | +12% |
| Total Orders | N | -3% |
| Average Order Value | $XX | +15% |
## Top [Category] by [Metric]
| Rank | [Category] | [Metric] | % of Total |
| ---- | ---------- | -------- | ---------- |
| 1 | [name] | [value] | XX% |
...
## [Section — e.g., "Regional Breakdown"]
[Table or narrative]
## Notable Findings
- **[Finding 1]:** [1–2 sentence explanation and implication]
- **[Finding 2]:** [1–2 sentence explanation and implication]
## Data Notes
- Source: [filename / system]
- Records analyzed: N
- Missing data: [Any caveats]
Step 5 — Tailor to Audience
| Audience | Style adjustments |
|---|
| Executive | Lead with money/growth metrics, minimize tables, add implications |
| Analyst | More data tables, include methodology notes |
| Customer-facing | Only relevant metrics, no internal identifiers, softer language |
| Internal team | More detail, include raw counts alongside percentages |
Output Format
Primary: Markdown report (copy into Notion, Confluence, or email).
Optional outputs:
- Email format: Subject line + 3-paragraph prose + bulleted highlights
- Slide outline: Title slide → key metrics → top findings → next steps
Safety & Confirmation
- Never share PII from the dataset (names, emails, phone numbers) in the report output.
- Flag any revenue or user count numbers before including — confirm the user wants them visible.
- Label estimates and inferences clearly: "Based on available data..." vs. confirmed facts.
