write-like-me

Write and edit prose in a specific author's voice across chat, reviews, docs, and blogs.

3|Updated Jun 17, 2026
One-click install
npx skills add https://github.com/ronaknnathani/relay --skill write-like-me-ronaknnathani
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: write-like-me
Source: https://github.com/ronaknnathani/relay/tree/main/skills/write-like-me
Command: npx skills add https://github.com/ronaknnathani/relay --skill write-like-me-ronaknnathani

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI-generated text often sounds generic and instantly recognizable as machine-written, which breaks trust when content goes out under a real person's name. This Skill encodes one author's concrete voice rules so drafts, edits, and replies read as if the person wrote them. ## Core Features & Use Cases - Voice rulebook: Enforces non-negotiable mechanics such as no em-dashes, no semicolons, mandatory contractions, plain words, and one idea per sentence. - Register selection: Picks the right surface style first, from lowercase terse chat and PR review comments to fully formed prose for design docs, PR descriptions, peer feedback, and blogs. - Editing mode: Tightens existing drafts by cutting repetition and filler while preserving the author's claims, structure, and technical model. - Use Case: You need a PR review comment, a design doc recommendation section, or a reflective blog paragraph that sounds like you rather than a generated reply. The Skill applies the signature phrases, confidence calibration, and pre-send checklist before anything ships under your name. ## Quick Start Ask the agent to rewrite this PR description in my voice using the write-like-me skill, keeping it concise and free of em-dashes.

Frequently Asked Questions about write-like-me

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make AI-generated writing sound like me?▼

Define explicit voice rules covering banned punctuation, sentence length, word choice, and signature phrases, then apply them per surface. This Skill encodes those rules so drafts and edits match the author's style across chat, reviews, docs, and blogs.

How to write PR review comments that don't sound AI-generated?▼

Keep review comments terse, lowercase, and Socratic, around 14 words with honest severity markers like nit: and q:. Avoid the acknowledge-plus-SHA-plus-mechanism cadence, which is a reliable generated-reply tell.

What are the biggest tells of AI-generated text?▼

The most reliable tells are em-dashes, semicolons, hype words like seamless and robust, padded template sections, and restating the diff in prose. Cutting these and grounding every claim in a name, number, or example removes most of the generated feel.

Can this skill edit existing drafts instead of writing new ones?▼

Yes, editing mode cuts repetition, improves flow, and removes banned punctuation while preserving the author's claims, idea order, and technical model. It makes text sound more like the person rather than rewriting it into an editor's voice.

Does the writing style change between chat and design docs?▼

Yes, the register table defines different rules per surface. Chat and code review are fast and lowercase, while docs, blogs, and feedback use proper capitalization and fully formed prose with recommendation-first structure.