posthog/writing-user-facing-copy
> Use ALWAYS before writing or editing user-facing copy, and whenever a code change adds or changes a string a user will see. Also carries feature-naming rules that are easy to get wrong, most importantly how to talk about the Wizard (name it "Wizard" once, then call it "the setup agent" / "the agent"). Not for internal code comments, commit messages, or variable names.
npx skills add https://github.com/PostHog/posthog --skill writing-user-facing-copy
This is the operational gate for anything a person reads in the product or around it.
Run it before writing or editing user-facing text, and whenever a code change introduces or changes a visible string.
Applies to: UI labels, buttons, form fields, tooltips, empty states, error and success messages, toasts and notifications, onboarding flows, emails, docs, and support replies.
Does not apply to: code comments, commit messages, log lines, variable/function names, or other developer-only text.
Write the way a person would. Neutral and humane.
surveyPopupDelaySeconds becomes "Delay the survey popup".Do not use em-dashes (—) anywhere in user-facing copy.
Do not substitute an en-dash either.
Prefer rewriting the sentence so the dash isn't needed. If a connector is unavoidable, use a hyphen with spaces, a comma, a colon, or split into two sentences.
Also avoid the sentence _shapes_ that lean on that dash, because they read as machine-written:
Write the plain version instead.
This is the failure mode that slips through even when you know the rule, so check for it specifically. It applies to every string, including headings, section titles, and button labels, not just body text.
The tells:
After drafting, re-read and cut anything that's there to sound good rather than to inform. If a line would feel out of place said plainly out loud to a colleague, rewrite it.
Say what happened and give the next action. Don't stop at reporting that something failed.
Use the names users see, and stay consistent across every surface.
The setup tool is named Wizard. It confuses users as a description, because "wizard" reads like an old-style step-by-step form, not an AI. Users already have a mental model for AI agents.
Examples:
For any other feature, use its product-facing name exactly as it appears in the UI, and describe it in the terms users already understand.
If you can't tell whether copy reads well, or whether a term is the right user-facing name, ask a human before shipping it.
Take posthog/writing-user-facing-copy from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.