Skip to content

Muvon/octomind-tap / skills/analytical-report

analytical-report

Verified

Turn raw data into a decision-ready report: answer-first Minto Pyramid, traced numbers, charts, HTML that prints to PDF. Use for data reports.

Overview

A report exists to help a human decide something. Most data reports fail two ways: they bury the answer under process (a chronological data dump), or they fabricate/round numbers the underlying data doesn't support. This skill encodes how to produce a report that leads with the answer, proves it with real numbers traced to the data, and is delivered in a form a person can actually read and print.

Mental model

Two non-negotiables: the pyramid (answer first) for readability, and the grounding gate (every number computed from the data) for trust. Everything else — chart choice, layout, print CSS — serves one of those two.

Instructions

Structure — the Minto Pyramid (answer first)

Executives read top-down and have little time; match that. Structure every report as a pyramid:

  1. Headline finding — the single most important takeaway, in one sentence, at the very top. If the reader stops here, they have the answer.
  2. Executive summary — 3–5 key findings (the rule of 3–5), strongest first, each one sentence with its supporting number. This is the whole report in a glance.
  3. Supporting sections — one per key finding, in the same order. Each leads with its finding (not a heading like "Data"), then shows the evidence: the chart, the table, the method in a line.
  4. Recommendations / next steps — what to do about it, when the report's purpose is decision-support.
  5. Appendix / method — data sources, date range, caveats, and how figures were computed, for the reader who wants to verify.

Never open with "This report analyzes…" or a methodology wall. Open with the finding.

Grounding gate — every number traced to the data

  • Compute every statistic from the actual data (load it, run the numbers — don't estimate). Report figures at the precision the data supports; don't invent decimals or round in a way that changes the story.
  • No fabricated or illustrative numbers. If a figure can't be computed from the provided data, say so — never fill the gap with a plausible-looking number.
  • State the data source, the date range/scope, and any filtering. Show the method in one line where a number could be questioned ("median of 1,204 sessions, bot traffic excluded").
  • Flag uncertainty honestly: small samples, missing data, outliers that move the result. Give a range or interval next to any estimate that a decision hinges on. A caveat is cheaper than a wrong decision.
  • Correlation is not causation — describe what the data shows, not a causal story it can't support.

Charts — the right one, labelled, self-contained

Pick the chart by the question, not by variety:

  • Trend over time -> line. Comparison across categories -> bar. Part-to-whole -> stacked bar (pie only for 2–4 slices, sparingly). Distribution -> histogram. Relationship -> scatter. A single number that matters -> a big bold stat, not a chart.
  • Label directly (values on/next to marks) over forcing the eye to a legend and axis. Title each chart with its takeaway, not its variable names ("Signups fell 40% after the April change", not "Signups by month").
  • One visual system: consistent colours, fonts, and sizing across every chart so the report reads as one document. Use colour to encode meaning, not decoration; keep it colour-blind-safe and readable in greyscale (it will be printed).
  • Render charts as inline SVG embedded in the HTML — self-contained, styleable by the print CSS, no external image files or CDN dependencies. Plot with matplotlib and the bundled house style (report.mplstyle, listed under Skill Resources): plt.style.use("<path>/report.mplstyle"), grey the context series and color the one that carries the message, ax.set_title(<takeaway>), ax.bar_label() or ax.annotate() for direct labels, then fig.savefig(buf, format="svg") and inline the SVG. Run it with uv run --with matplotlib python chart.py when matplotlib isn't installed. Hand-written SVG only for sparklines and single bars; drawn-by-hand axes and labels come out misaligned.
  • Make every chart readable without sight: role="img" plus a <title> (and <desc> for the takeaway) inside the SVG, and the underlying numbers available as a table nearby or in the appendix. Keep marks at 3:1 contrast against the background and never rely on colour alone — add labels, patterns, or direct annotation.

Delivery — self-contained HTML that prints to clean PDF

Default output is a single self-contained HTML file: all CSS inline in a <style> block, all charts inline SVG, web-safe or embedded fonts — no external requests. It opens in any browser (human-readable immediately) and prints to PDF via the browser (Print -> Save as PDF) with no tools installed.

Make it print cleanly:

@media print {
  @page { size: A4; margin: 18mm; }
  .no-print { display: none; }
  h2, h3 { break-after: avoid; }
  table, figure, .card { break-inside: avoid; }
  a { color: inherit; text-decoration: none; }
}
body { font: 11pt/1.5 -apple-system, "Segoe UI", Roboto, Helvetica, Arial, sans-serif; color: #1a1a1a; }
td.num, .stat { font-variant-numeric: tabular-nums; text-align: right; }

Use absolute units (pt, mm) for print layout, break-inside: avoid on tables/figures/cards (the older page-break-* properties are legacy aliases), and a print-only footer (page context, source, date). Keep the on-screen version clean too — this is one document for both. Set <html lang> and a real <title>, and use semantic headings and <table> with <th scope> so the HTML and the PDF exported from it stay accessible.

Typography and layout for humans: generous line height, a readable measure (not full-width text), clear heading hierarchy, tables for enumerable facts with right-aligned numbers, a stat-tile row for the headline metrics, whitespace over density. Restraint reads as authoritative.

Format flexibility: HTML is the default and the most useful (browser + PDF in one). Produce Markdown if the request asks for it, or a rendered PDF if a renderer is available in the environment (the agent can render the HTML via a shell tool when one is present); otherwise deliver the HTML and tell the reader to open it and Print -> Save as PDF.

Examples

Weak opening (process-first, no answer): "This report presents an analysis of user engagement data collected between January and June 2026 across all platform surfaces, using the following methodology…"

Strong opening (answer-first, pyramid): "Engagement is up 22% — but entirely from existing power users; new-user activation actually fell 15%, and that's the risk. The three findings below, then the detail."

Chart title, weak: "Sessions by Month". Chart title, strong: "Sessions doubled after launch, then plateaued in Q2 — the growth engine has stalled."

Checklist

Before finalizing:

  • The first sentence gives the answer.
  • Every key finding is readable from the executive summary alone.
  • Every number traces to the data, with its method, at the precision the data supports; nothing is illustrative or estimated.
  • Every chart is the right type, directly labelled, titled with its takeaway, part of one visual system, and has a text alternative.
  • Caveats name sample size, missing data, and anything correlational.
  • The HTML renders standalone with no external requests and prints without broken tables or clipped charts.
canvas-designanthropics/skillsCreate beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
docxanthropics/skillsUse this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx files) or Word templates (.dotx files). Triggers include: any mention of 'Word doc', 'word document', '.docx', '.dotx', or requests to produce professional documents with formatting like tables of contents, headings, page numbers, or letterheads. Also use when extracting or reorganizing content from .docx or .dotx files, inserting or replacing images in documents, performing find-and-replace in Word files, working with tracked changes or comments, or converting content into a polished Word document. If the user asks for a 'report', 'memo', 'letter', 'template', or similar deliverable as a Word or .docx file, use this skill. Do NOT use for PDFs, spreadsheets, Google Docs, or general coding tasks unrelated to document generation.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
pdfanthropics/skillsUse this skill whenever the user wants to do anything with PDF files. This includes reading or extracting text/tables from PDFs, combining or merging multiple PDFs into one, splitting PDFs apart, rotating pages, adding watermarks, creating new PDFs, filling PDF forms, encrypting/decrypting PDFs, extracting images, and OCR on scanned PDFs to make them searchable. If the user mentions a .pdf file or asks to produce one, use this skill.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
pptxanthropics/skillsUse this skill any time a .pptx or .potx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx or .potx file (even if the extracted content will be used elsewhere, like in an email or summary); editing, modifying, or updating existing presentations; combining or splitting slide files; working with templates (.potx), layouts, speaker notes, or comments. Trigger whenever the user mentions "deck," "slides," "presentation," or references a .pptx or .potx filename, regardless of what they plan to do with the content afterward. If a .pptx or .potx file needs to be opened, created, or touched, use this skill.Stars: 198kUpdated Claude CodeCodexVerifiedInstall