technical-writing
Skill0 starsChanged 51 days agoMITView on GitHub
Structure and plain sentences for writing an engineer acts on, end-of-task reports, PR descriptions, commit messages, docs, RFCs. Use when drafting or reviewing any of these, including your own report at the end of a task. Pair with clarity, which owns voice.
142 other projects use a version of this, 1 of them edited.
Adopted by
Most recently taken up by
- Yeachan-Heo/oh-my-claude-sisyphustheir own version, first seen Nov 8, 2026
- Loner1024/skillstheir own version, first seen Nov 8, 2026
- lklimek/claudiustheir own version, first seen Nov 8, 2026
- justinyoo/skillstheir own version, first seen Nov 8, 2026
- thrillmade/agent-skillstheir own version, first seen Nov 8, 2026
- metraton/gaiatheir own version, first seen Nov 8, 2026
Used with
Projects with it also use ketchup-video, principle-explain-the-number and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
tristanphvnSkillketchup-video
Catch the user up like /keel:ketchup, with a short silent video of slides for what happened and the waiting calls as text under it. Use for /keel:ketchup-video.
Also in 14 projects that use this
ianclearySkillprinciple-explain-the-number
Apply before you trust, report, or act on a number you measured: a speedup, a regression, a throughput, a latency, or an eval result. Find what limits it, and rule out that it measured something other than the work you think.
Also in 29 projects that use this
eryzerzSkillmake-bot-ui
Use when building a custom UI (page, dashboard, buttons) that should wake a Grok Bot over a webhook, when the user must provide a webhook sender key, or when exposing that UI on Tailscale.
Also in 20 projects that use this
jaydubya818Skillprinciple-attack-the-premise
Apply when repeated fixes fail and a shared assumption may be wrong.
Also in 35 projects that use this
jaydubya818Skillprinciple-test-behavior-not-implementation
Apply when tests mirror internals instead of asserting observable outcomes.
Also in 36 projects that use this
ianclearySkillpoteto-help
Guides users through pstack setup, /poteto-mode, and picking the skill, playbook, or principle for a task. Type /poteto-help with a question.
Also in 14 projects that use this
- +5 more from jaydubya818
Other versions142 projects use a version of this.
Word for word · 0
None found.
Edited · 1
- endqwerty/endstack★ 111% same
Wrote their own version · 141
- Yeachan-Heo/oh-my-claude-sisyphus★ 41ksame name
- Yeachan-Heo/oh-my-claudecode★ 41ksame name
- vercel/eve★ 5.8ksame name
- backnotprop/pstack★ 3.8ksame name
- chattocorp/chatto★ 2.9ksame name
- cashew-labs/libretto★ 898same name
- saffron-health/libretto★ 883same name
- moeru-ai/xsai★ 667same name
- proffesor-for-testing/agentic-qe★ 524same name
- NPC-Worldwide/npcsh★ 502same name
131 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
7 rules. 0 are common, 7 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 134 edited versions we read keep these.
7
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
- Distinctive: line 29
Leave out: the narrative of how you got there, files read, dead ends, restating the request, a closing offer of more help.
- Distinctive: line 44
Don't mix modes in one document.
- Distinctive: line 61
Keep "only" and "not" next to the word they change.
- Distinctive: line 63
Every "it", "this" and "they" points at one obvious thing.
- Distinctive: line 64
Don't drop verbs: "Phase 1 moves the converters and Phase 2 the runtime" leaves Phase 2 without one.
- Distinctive: line 65
Call each thing by one name everywhere.
- Distinctive: line 66
No slashes for "and/or": write "a, b, or both".
“Common” is measured: kept by at least 6 in 10 of the 134 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.
Files
Technical writing
The goal: a tired engineer understands it on the first read and knows what to do next.
Three rules sit above everything else:
- Cut every word that does no work. "In order to" is "to". "It is important to note that" is nothing.
- Use the short, everyday word. "Use", not "utilize". "Do", not "perform".
- When a rule makes a sentence worse, fix the sentence another way or leave it. The rules serve the reader. A sentence that obeys every rule and reads like a machine wrote it has failed.
The codebase is the word list. Write the real symbol, file, flag or command, not a description of it. Don't invent jargon; say what a developer would say out loud.
Reports and PR descriptions
These are the writing an agent produces most, and where length costs the most.
End-of-task report to the user. In this order, and nothing else:
- Outcome — one or two sentences: what changed and where (link the PR).
- What the reader must know — decisions made on their behalf, behaviour that changed, surprises. Only what changes what they do or believe.
- Checks — what ran and passed, what was deferred, what was not verified. A deferred check is stated, never implied to have passed.
- What they need to do — merges, approvals, commands to run. Numbered if more than one.
Leave out: the narrative of how you got there, files read, dead ends, restating the request, a closing offer of more help. If a detail only matters to someone reviewing the diff, it belongs in the PR, not the report.
PR description. A briefing a reviewer reads in under a minute: why, what changed (grouped by area, not by file), how it was verified, what was deferred, anything the reviewer must decide. Link logs and long tables instead of pasting them. Follow the repo's PR template when one exists.
Commit message. Subject says what the commit does. Body says why, and anything non-obvious a future reader of git log needs.
Pick the mode first (longer docs)
One document, one mode. Does the reader need to do something or understand something, and are they learning or working?
- Tutorial (doing, learning): says what the reader will build; every step produces a visible result; explanation cut to a clause and a link.
- How-to (doing, working): solves one problem the reader has; assumes competence; action only. Name it by the task.
- Reference (understanding, working): facts, options, limits, errors. Dry, complete, no opinion. Mirrors the structure of the thing it describes.
- Explanation (understanding, learning): one topic, the why — decisions, constraints, alternatives. Opinion belongs here and nowhere else.
Don't mix modes in one document. Split and link.
Sentences
- Address the reader as "you", in the present tense.
- Say who does what: "the guard rejects", not "is rejected". Passive only when the actor is unknown or irrelevant.
- Instructions are commands: "Run
pnpm build." Put the condition first: "To re-vendor, run …". - Common case first, exceptions after.
- One instruction per sentence; one thought per sentence elsewhere. Split sentences past about 25 words — but keep a long sentence that carries one thought with its condition.
- Vary length on purpose. All-short reads clipped and mechanical.
- Be specific: not "schema changes can cause issues" but "a column rename fails the build".
- Headings carry the point ("Build before you test"), in sentence case.
- Numbered lists for sequences, bullets otherwise. Introduce a list with a sentence; keep items parallel.
- Code in code font. No "simply", "easy" or "just" in instructions.
No sentence with two readings
- Keep "only" and "not" next to the word they change.
- Break up long noun strings: "the vendor tarball hash verification step" → "the step that checks the tarball hashes".
- Every "it", "this" and "they" points at one obvious thing. Repeat the noun when in doubt.
- Don't drop verbs: "Phase 1 moves the converters and Phase 2 the runtime" leaves Phase 2 without one.
- Call each thing by one name everywhere. Three names for one thing teaches three things.
- No slashes for "and/or": write "a, b, or both". No "(s)" plurals.
- Skip idioms, Latin abbreviations and metaphors. Non-native readers and agents parse plain constructions best.
- Punctuation (em dashes, semicolons, parentheses) is judged by frequency and fit, per
clarity, not banned.
Worked example
Before:
I went ahead and took a look at the audit service and it turned out that there were a number of places where the diff was being computed manually, so I refactored those to use the new helper, which should make things more consistent going forward. I also ran the tests and they seem to be passing.
After:
Audit entries now go through
logChange; the six hand-built diffs are gone. Tests:audit,rolesandpatientsspecs pass (304). Not run: the web app in a browser.
Source
Reworked from pstack's technical-writing (Lauren Tan, MIT), which layers Diátaxis, Google developer style, ASD-STE100 and Global English. Changes: model-invoked instead of manual; a section for agent reports; punctuation deferred to clarity rather than banned; unslop replaced by clarity. See source-reference.md.
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.

