← ClaudeAtlas

voice-builderlisted

Ingests 5–20 of the user's writing samples and distils a reusable voice-profile.md — signature phrases, sentence-length distribution, opener/closer habits, punctuation quirks, vocabulary do/don't lists, tone sliders, and three calibration paragraphs with self-checks. Other skills load this file so every post sounds like the user, not like AI. Use when the user says "make it sound like me", "clone my writing voice", "everything you write sounds like AI", or "build my voice profile".
alebgl77/claude-inc · ★ 9 · AI & Automation · score 77
Install: claude install-skill alebgl77/claude-inc
# Voice Builder — Voice Coach > "Clone your voice" ## When to use - "Everything you write sounds like AI — make it sound like me." - Before any recurring content work: build the profile once, reuse it everywhere. - "Clone my voice from these posts." / "Here are my last 15 newsletters." - Ghostwriting for a founder whose voice must survive the ghostwriter. - Recalibrating after feedback: "that draft didn't sound like me." ## Workflow 1. Collect 5–20 samples of the user's own unedited writing — posts, emails, DMs, talk transcripts. Reject copy written by others or by AI. Under five usable samples: proceed, but stamp the profile low-confidence and name what's missing. 2. Run the mechanical analysis: sentence-length distribution (short / medium / long %), paragraph size, punctuation quirks (em-dashes, ellipses, parentheticals, one-word sentences), capitalisation habits, emoji policy. 3. Extract the lexical fingerprint: signature phrases and crutch words, a DO list (words they reach for), a DON'T list (words they would never use), slang and profanity tolerance. 4. Map the habits: how they open (question, scene, blunt claim), how they close (CTA, punchline, trail-off), how they handle transitions and humour. 5. Set the tone sliders — formal↔casual, warm↔dry, confident↔hedged, dense↔airy — each a 1–10 position justified by a quoted sample line. 6. Write three calibration paragraphs on neutral topics in the reconstructed voice, each with a self-check: