Ghostwriter
Teach your agent how to write like you. A portable two-mode skill that turns your sent writing into a style guide your agent uses for drafts that sound like you, not like AI.
Install
npx skills add joshferrara/ghostwriterInstall command taken from the project's README or registry entry. Always review code before running it.
README excerpt
Read full README on GitHub →Ghostwriter
Teach your agent how to write like you.
Your agent already writes well. It just doesn't write like you. Ghostwriter fixes that. It studies the messages you've actually sent and turns them into a style guide your agent uses to draft in your voice.
Generic AI You, via Ghostwriter
────────── ────────────────────
"I hope this email finds you "hey, quick one. did the
well! I wanted to circle back contract ever get signed?
regarding the contract. Please no rush, just closing tabs."
let me know at your earliest
convenience. Thanks so much!"
Same intent. One of them is yours. Once your guide is in place, agent-written drafts should feel spooky good.
Quickstart
# 1. Install (the skills CLI puts it in the right place for your agent)
npx skills add joshferrara/ghostwriter
# ...or clone by hand (one clone covers most agents; see "Install" below)
git clone https://github.com/joshferrara/ghostwriter.git ~/.claude/skills/ghostwriter
# 2. Build your guide (just talk to your agent)
"Build my Ghostwriter guide from my sent email."
# 3. Use it
"Draft a reply to this client in my voice."
That's it. Step 2 walks you through picking a source and confirming privacy; step 3 reads the guide and drafts. The rest of this README is the details behind those three lines.
Why
Most AI writing has the same tells: inflated significance, "I hope this helps!", forced rule-of-three, and a flat, sourceless cadence. You can spot it in a second, and so can everyone you write to.
Generic "humanizing" advice strips those tells, but it pushes everything toward the same neutral human voice. That's not you either. You have specific habits: the way you open, the sign-off you actually use, whether you reach for em dashes or emoji, how blunt you get when you're in a hurry.
Ghostwriter inverts the priority. It learns your voice from your sent messages and writes that first. Only then does it run a light cleanup pass for AI artifacts, and that pass never overwrites what makes your writing yours.
How it works
Point it at your sent messages → it writes a style guide → your agent drafts as you. Two modes do the work:
- Generate mode analyzes writing you've actually sent (email, texts, Slack, exports) and produces a standalone style guide, usually
ghostwriter-guide.md. - Use mode loads that guide and drafts or edits in your voice, treating your personal style as the primary authority.
GENERATE USE
───────── ───
detect existing access ─┐ find your guide
select source(s) │ read it
confirm privacy ├─► ghostwriter- gather missing context
collect SENT messages │ guide.md ──► draft in your voice
analyze 17 dimensions │ humanizer cleanup (secondary)
write the guide ─┘ deliver
Generate mode
Triggers when you ask to create, generate, build, update, refresh, or improve your Ghostwriter guide. The agent walks you through:
Similar plugs
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local…
Why use many token when few token do trick. Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.
Production-grade engineering skills for AI coding agents.
The 1st Collaborative AI Design Agent Workspace. Local-first and open source. Create prototypes, websites, apps, dashboards, presentations, images, and videos with your team…