# Progressive Reading

> Make an answer easier to start, scan, pause, and resume without making it shallower: a reply that is correct but too dense to enter, a wall of text with the useful part buried, a summary that dropped the caveat that mattered, or prose that reads as machine-generated. Covers useful-answer-first ordering, short paragraphs, literal headings, preserved nuance, and a small diagram where structure is the difficulty. Use when the user says "I do not understand", "explain this more simply", "make this clearer", "too dense", "break it into parts", "step by step", or "without losing the details". Not for shortening at the cost of accuracy, not for compressing an ordered procedure or a security warning, and not for a draft that sounds generated rather than dense, which is a question about voice and belongs elsewhere. An answer nobody can get into was not delivered, and one shortened until it is wrong was not either.

- Skill: `gabrielmoreira/progressive-reading` (Agent Skill, multi-file: 6 files)
- Install (CLI): `npx skillmds@latest add gabrielmoreira/progressive-reading`
- Raw SKILL.md: https://api.skillmd.com/api/skills/gabrielmoreira/progressive-reading/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: gabrielmoreira (https://skillmd.com/u/gabrielmoreira)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/gabrielmoreira/progressive-reading

---


# Progressive Reading

**Core principle.** The reader can start, scan, pause, resume, and understand, without losing what matters.

- **Lower the effort to read, never the correctness or the depth.** Cutting substance is not simplifying. It is losing the answer.
- **You opened this in the middle of something.** This is how to do that work, not a replacement for it. Name what you were doing before you start, and return to it when this is done.

## Order and structure

- **Lead with the useful answer.** Context comes after it, caveats after that.
- **One main idea per short paragraph.** Split a dense explanation into chunks the reader can stop between.
- **Staccato is the failure on the other side of that rule.** A short paragraph is a place to stop, not a target to hit. Two ideas that depend on each other belong in one sentence, where the dependency is visible; split apart they become two facts and no argument. A page where every paragraph runs a single line and every line opens in bold has traded its connective tissue for scannability, and can be entered anywhere but followed nowhere.
- **The check is whether the joins survive.** Read three consecutive paragraphs and ask what carries one to the next. If the answer is only that they sit in order, they were cut, not structured.
- **Use a heading only where it helps navigation.** Open each section with its point.
- **Keep related information together.** A short summary goes before any dense detail it introduces.
- **Reach for a bullet, a checklist, or an example only when it scans better than a sentence.** Never build a wall of bullets.
- **Rank a long list, never truncate it.** Eight findings ordered by what matters lose nothing; a cap at five loses three. The number of items is the reader's problem to skim, and the order is yours to solve.
- **Keep spacing compact and steady.** Blank lines, forced symmetry, and decorative structure make an answer longer, not clearer.

## What the reader cannot hold

- **Anything not on screen is gone.** Never write "keep in mind", "as established above", or "recall that". Where a fact is needed here, put it here, and pay the repetition.
- **A pointer backwards costs a scroll and usually a re-read.** Restating the thing in four words is cheaper than sending the reader to find it.
- **Between turns even less survives.** Where work runs across replies, open by saying where it stands: what is done, what this turn changes, what is left.

## Understood is not acted on

- **The gap between knowing and doing is where an answer dies.** A reply that is correct, readable, and leaves the reader working out what to do with it has not landed.
- **End on what the reader does next**, where there is one. A command, a file to open, a decision that is theirs. Not a summary of what was just said.
- **Where nothing is next, stop.** An invented next step costs more than an absent one.

## Voice

- **Be human, direct, and specific.** Simple verbs and natural rhythm beat inflated phrasing.
- **Cut the packaging.** All of it goes.
  - Empty praise.
  - Boilerplate openings.
  - Points already made.
  - Generic conclusions.
  - "Let me know if" endings.
- **Avoid corporate, salesy, or artificially polished language.** Typographic gimmicks belong in none of it.
- **A draft that sounds generated rather than dense is a different job.** Voice, over-claiming, and the shape a message arrives in belong to the skill that owns them. This one owns pace, order, and entry.
- **State uncertainty only where it is real.** Never invent a fact, an example, a citation, or a source.

## Preserve meaning

- **Cut filler. Never cut substance.**
- **Keep these exactly as written**, unless the user asked for them to change.
  - Technical terms and identifiers the domain already uses.
  - Code, commands, and paths.
  - API names.
  - Quoted error strings.
  - A term coined in the draft to compress a decision is voice, not a term, and belongs to the skill that owns voice.
- **Preserve what a shorter version would quietly drop.**
  - Nuance.
  - Risks and tradeoffs.
  - Exceptions.
  - The edge cases that carry weight.
- **Never shorten these where compression could blur meaning, order, or risk.**
  - A security warning.
  - A confirmation before an irreversible action.
  - An ordered procedure.
- **Stop before the answer turns wrong, misleading, or too terse to follow.**

## ASCII diagrams

- **Use one only where it beats prose** at showing one of these.
  - Structure.
  - Flow.
  - Ownership.
  - Relationships and tradeoffs.
- **Keep it simple.** A decorative diagram costs the reader attention and returns nothing.

## Before returning

- **The useful answer comes first.**
- **It can be scanned.**
- **It is technically complete.**
- **It reads as continuous prose where the reasoning is continuous.** Not "it reads naturally", that is unfalsifiable, and next to a scannability check that is concrete it always loses. Point at the sentences that carry the argument between paragraphs.
- **Nothing in it is unsupported.**
- **Read only the first line and the last line.** Together they should carry what to do and what happened. If they do not, the useful parts are buried in the middle.

