agentleFS
Smirnov-LabsSmirnov-Labs/claude-skills/skills/minto-pyramid/SKILL.md

minto-pyramid

Skill6 starsChanged 9 days agoMITView on GitHub

Use when tightening, shortening, cleaning up, restructuring, or de-waffling a draft — email, memo, status update, PR description, proposal, or announcement — especially when it opens with context instead of the conclusion, buries the recommendation or ask at the end, reads as a chronological recap, rambles, or needs to be exec-ready or compressed into a short summary. Also when the writer asks what point they're actually making, or whether they've buried the lede.

5 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Surface, don't soften: If the real conclusion is a hedge ("I'm leaning toward…"), keep the hedge — but still put it first.

    line 76

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Minto Pyramid

Overview

Lead with the answer. Barbara Minto's Pyramid Principle (developed at McKinsey): state your single conclusion first, support it with exactly three mutually-exclusive reasons, then put evidence under each. Most drafts do the opposite — they walk through context chronologically and reveal the point at the end.

Core principle: The reader should know your conclusion by the end of the first sentence. Everything after it exists only to support it.

When to Use

Apply when a draft:

  • Opens with background or context instead of the conclusion
  • Buries the recommendation, ask, or decision in the last paragraph
  • Reads as a chronological recap ("we did X, then Y, then Z…")
  • Organizes into topical buckets ("what's working / what's not / thoughts") with the ask at the bottom
  • Rambles, over-explains, or mixes the point with nice-to-have detail
  • Needs to be exec-ready, skimmable, or compressed into a short email

Also when the writer asks "what's my actual point?" or "did I bury the lede?"

Skip it for: pure narrative (stories, post-mortems where sequence matters), reference docs, or anything where suspense or chronology is deliberately the point.

The Pyramid

              [ Single conclusion ]          one sentence, first
             /         |          \
        Reason 1    Reason 2    Reason 3     exactly three, MECE
        /  |  \      /  |  \      /  |  \
       evidence     evidence     evidence    facts — only what earns its place

MECE = Mutually Exclusive (the reasons don't overlap) and Collectively Exhaustive (together they cover the case).

The Five Moves

Apply in order. Each is a self-contained pass over the draft.

#MoveWhat it does
1RestructureRebuild answer-first: conclusion → 3 reasons → evidence. Cut anything that doesn't sit under one reason.
2Buried-lede testFind the ONE sentence the draft is really building toward. If it isn't at the top, move it there.
3Three-reason auditDo the reasons overlap? Do they together cover the case? Is any one a detail pretending to be a reason?
4So-what passRead each line and ask "so what?" Delete every line that can't answer it.
5Email versionCompress to <150 words, same structure, evidence only where it earns its place. Keep the author's voice.

Guardrails

  • Lead with the ask. Even topical buckets ("what's working / what's not / recommendation") strand the decision at the bottom — put it first.
  • Exactly three reasons. Two feels thin; four means one isn't load-bearing or two overlap. Audit and merge until three load-bearing reasons remain.
  • Evidence earns its place. Credit, backstory, and process detail usually fail the "so what?" test for a decision-maker. Cut them, or demote to one line.
  • Preserve voice. Compress and reorder; don't rewrite the person into a robot. Keep their phrasing, hedges, and tone. Answer-first ≠ generic.
  • Surface, don't soften. If the real conclusion is a hedge ("I'm leaning toward…"), keep the hedge — but still put it first.

Example

Before (buries the lede — the ask is the last sentence):

Hey all — quick update on the onboarding wizard. We rolled it out to half of signups two weeks ago. Activation is up ~4% but that's within noise. Time-to-first-project dropped 9→6 min. Support's seeing ~20 extra mobile tickets/week and step 2 has 30% drop-off. The pricing redesign is specced and might move conversion more. I'm mostly thinking we pause the wizard and move to pricing.

After (answer first):

My lean: pause the wizard rollout and move the team to pricing next cycle. Three reasons:

  • Mobile is hurting users — ~20 extra support tickets/week, and step 2 drops 30%.
  • The fix is expensive — ~3 more weeks of eng to get mobile and step 2 right.
  • Pricing is the better bet — fully specced and likely to move conversion more.

Upside so far is thin: time-to-first-project dropped 9→6 min, but activation is only up ~4% (within noise). Curious what people think.

The conclusion moved from the last sentence to the first. Three MECE reasons. The so-what pass cut the rollout date and demoted the weak activation stat to one honest caveat. Voice intact.

The Five Prompts (verbatim — to run by hand)

The moves above as copy-paste prompts, if you'd rather drive them one at a time:

  1. Restructure — "Rewrite this using the Minto Pyramid: open with my single main conclusion in one sentence, then exactly three supporting reasons, then the evidence under each. Cut anything that does not sit under one of the three."
  2. Buried-lede test — "Read this draft and tell me the one conclusion it is actually building toward. If I only had one sentence to give my boss, what would it be? Then tell me where in the draft I hid it."
  3. Three-reason audit — "Check my three reasons. Do they overlap, and together do they fully cover the case? Flag any that is a detail pretending to be a reason."
  4. So-what pass — "Go line by line and ask so what after each one. Delete every line that cannot answer it. Show me what survives."
  5. Email version — "Compress this into an email under 150 words in the same structure: answer, three reasons, evidence only where it earns its place. Keep my voice."

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.