write-in-your-own-voicelisted

# Write in your own voice ## Purpose A maintainer reads a lot of PR bodies and issue text, and the generic-LLM register is instantly recognizable: "comprehensive", "robust", "leverages", "seamless", "this PR addresses the issue where…", and paragraphs of hedged padding around a one-line change. That register reads as slop before the reviewer has looked at a single line of the diff, and it costs you trust you have not spent yet. Prose in the house's own plain voice does the opposite — it signals a human who did the work and respects the reader's time. This governs **human-facing prose** — the register of the text you write to people. It is distinct from the sibling rule about source-code comments: the AI tell shows up on two different surfaces, and this skill fights it in the prose. ## When to use - When drafting a PR body, an issue, or a review reply on a repo you don't own. - Any time a draft starts with "This PR addresses the issue where…" or leans on "comprehensive / robust / leverages / seamless / delve / it's worth noting". - When you want a voice sample: pull your own merged PRs to calibrate before writing. ## When NOT to use This is about human-facing PROSE; for whether an in-CODE comment should exist at all, that is comments-about-code-not-change. ## The practice (checklist) - [ ] Cut the AI-jargon words: comprehensive, robust, leverages, seamless, delve, "it's worth noting", "this PR addresses the issue where…". - [ ] Write what changed in one plain