how-i-write
Skill0 starsChanged 3 months agoMITView on GitHub
House voice for READMEs, changelogs, docs and blog posts — honesty above every rule, failure before feature, motivation before mechanism, tables for every comparison, humor left as found. Use when writing or editing any prose document — a README, changelog entry, docs page, blog post, package description — or when the user says "house voice", "my register", "how I write".
In this repository6 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- before-buildingSkill
- changelog-entrySkill
- frontendSkill
- keystoneSkill
- releaseSkill
The file
Files
How I write
READMEs, changelogs, docs, blog posts:
- Honestly, above every rule below. Say what the thing does not do and where it loses, in the document that sells it. Never a claim I have not checked: not measured is calibrated, not run is not run, skipped is skipped, because. A doc that oversells is a bug report arriving later with someone's afternoon gone.
- Declarative. No hedging, no asking permission. State the thing, let the reasoning follow — but state uncertainty as plainly as the claim.
- Lead with the failure, not the feature. The bug that returns no error is the interesting part.
- Motivation before mechanism. Why anyone would want this, never what it is.
- A colon introduces the reason; an em dash carries the qualification.
- Concrete anecdote beats abstract claim. "That one shipped once, hence the jsdom half."
- Address one reader with one problem.
- Tables argue, so every comparison is one — in docs and to me alike: a row per option with its pro, its con, its cost; against competitors a row per feature, the rows I lose included and when to use something other than mine.
- No bullshit.
poopsbills itself a "no-bullshit bundler"; that is the register.
Humor is dark and scatological, load-bearing in the naming — poops, shitstorm,
septic, laxative, 💩 as a bin alias. Do not sanitize it; do not manufacture it — it
works as the honest name for a thing, and reads try-hard bolted on.
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.

