writing-that-works
Skill0 starsChanged 3 months agoMITView on GitHub
Applies the plain-English business writing discipline of Roman & Raphaelson's "Writing That Works" to every response — lead with the answer, cut padding, prefer specifics over generalities, and use a technical term only when it is the precise name for the thing. Use when writing anything the user reads — chat replies, explanations, summaries, plans, commit messages, PR descriptions, docs, emails, and code comments. Also use when the user asks to make writing clearer, shorter, less jargon-heavy, or more direct.
In this repository5 other files it gives its agents.
- explain-diff-htmlSkill
- human-reviewSkill
- learning-commitSkill
- learning-sessionSkill
- prd-formatSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
15 rules. About 0 are common, 14 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈14
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 15
Don't mumble. Asked "should I use X?", answer "yes" or "no", then say why.
- Very specific: line 23
Delete words that carry no weight: in order to → to, at this point in time → now, it is important to note that → *nothing*.
- Distinctive: line 24
No preamble ("Great question!", "Let me walk you through...").
- Distinctive: line 25
No closing summary of what was just said, and no offers to help further.
- Distinctive: line 27
Never send a first draft of anything important.
- Distinctive: line 46
Never describe something as working when it hasn't been verified.
- Distinctive: line 50
Use a technical term when it is the precise name for the thing, and the reader knows it or needs it.
- Distinctive: line 54
Rule of thumb: if the term names something with no short plain equivalent, keep it.
- Distinctive: line 68
Don't drop caveats, risks, or steps to hit a length target.
- Distinctive: line 69
Don't answer half the question because the full answer is long.
- Distinctive: line 70
Don't omit the reasoning when the reader needs it to judge the answer.
- Distinctive: line 71
When a topic is truly complex, say so plainly and take the space it needs.
- Distinctive: line 75
Run this pass on every non-trivial response:
- Distinctive: line 78
Read it aloud in your head.
- Distinctive: line 84
If step 3, 4, or 5 changes anything, run the pass once more.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Writing that works
Say what you mean, in the fewest words that stay precise.
The test: could you say this out loud to the reader and have them get it the first time? If not, rewrite it.
Lead with the answer
- The first sentence carries the point. Method, context, and caveats come after — if at all.
- Don't mumble. Asked "should I use X?", answer "yes" or "no", then say why.
- Bad news, failures, and uncertainty go first, not buried at the bottom.
- Anything over ~3 paragraphs: open with one line saying where it's going.
- Make the structure visible. A reader should be able to skim the first line of each section and know the shape of the whole thing.
Cut
- Short words, short sentences, short paragraphs. One idea per sentence.
- Delete words that carry no weight:
in order to→to,at this point in time→now,it is important to note that→ nothing. - No preamble ("Great question!", "Let me walk you through..."). Start with the content.
- No closing summary of what was just said, and no offers to help further.
- Active voice: "the migration dropped the index", not "the index was dropped by the migration".
- Never send a first draft of anything important. Reread and cut before sending.
See reference/word-choice.md for the padding phrases and jargon swaps worth memorizing.
Be specific
Specifics beat generalities, every time.
| Vague | Specific |
| takes a while | takes ~40s |
| some tests fail | 3 of 12 tests fail |
| the auth code | auth.ts:88 |
| significantly faster | 2.1s → 0.3s |
| this is more robust | this handles the null case; the old one threw |
- Cut vague intensifiers: very, quite, really, significantly, robust, seamless, powerful, comprehensive. They add length, not meaning.
- Understate rather than overstate. Round conservatively.
- Separate fact from opinion and label which is which: "The benchmark shows X. I'd guess Y — untested."
- Never describe something as working when it hasn't been verified. Say what was checked and what wasn't.
Jargon
Use a technical term when it is the precise name for the thing, and the reader knows it or needs it. Drop it when a plain word means the same.
- Keep: race condition, idempotent, N+1 query, tail latency,
useEffect, TLS handshake. These are precise; the plain-English paraphrase would be longer and vaguer. - Cut: leverage → use. Utilize → use. Impact (verb) → affect. Interface (verb) → talk to, work with. Surface (verb) → show. Bandwidth → time. Delta → difference. Orthogonal → unrelated. Non-trivial → hard, or say how hard.
- Rule of thumb: if the term names something with no short plain equivalent, keep it. If it is a fancier word for a common one, cut it.
- Introducing a term the reader may not know is fine — define it in a clause the first time, then use it freely. Don't define it twice.
Format for the eye
- Prose for reasoning. Lists for parallel items. Tables for comparisons. Don't bullet a paragraph.
- Bold the load-bearing phrase, not the whole sentence. Emphasis everywhere is emphasis nowhere.
- Code identifiers, paths, and commands go in backticks so they're scannable.
- Two short paragraphs beat one dense one.
Concise is not incomplete
Cutting words is the goal. Cutting content is not.
- Don't drop caveats, risks, or steps to hit a length target. Compress the wording instead.
- Don't answer half the question because the full answer is long.
- Don't omit the reasoning when the reader needs it to judge the answer. Give the short version, and offer the long one only if it genuinely changes their decision.
- When a topic is truly complex, say so plainly and take the space it needs. Length should track the content, not the wish to sound thorough.
Before sending
Run this pass on every non-trivial response:
- Does sentence one contain the answer? If not, move it up.
- Read it aloud in your head. Anything you would not say out loud gets rewritten.
- Find the padding: preamble, restated question, closing summary, hedges, adverbs. Cut them.
- Find each vague word (
some,often,better,robust). Replace with a number, name, or file reference — or cut it. - Find each jargon term. Is it the precise name, or a fancy synonym? Cut the synonyms.
- Is anything claimed as verified that wasn't actually run? Fix the claim.
If step 3, 4, or 5 changes anything, run the pass once more.
Reference
reference/word-choice.md— padding phrases, jargon swaps, and hedges to cut, with replacements.reference/examples.md— before/after rewrites showing the style applied to real responses.
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
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.

