Build a voice guide from the writing you already have
Turn authorized writing samples into channel-specific voice rules and a shareable drafting prompt.
Where it came from: An August 2026 study produced a private voice guide, source audit and shareable prompt from cleaned and attributed writing, observed public material, preferences and one narrowly supported meeting register.
Package status: Draft package; structurally checked; not verified in Claude.
You give it
- Authorized writing samples
- Intended channels and audiences
- Authorship and assistance context
- Current writing preferences
You get back
- Core voice description
- Register cards
- Shareable drafting prompt
- Private source and uncertainty ledger
Most people asking AI to write in their voice hand it three recent posts. The model copies the short paragraphs, repeats a favorite phrase and turns every subject into the same performance.
The missing piece is context. An email asking for a refund, a comment on a colleague's post and a business article serve different jobs. A useful voice guide explains those differences so the next draft can make the right choice.
We have already done the underlying work. A voice study completed in August 2026 used cleaned, attributed writing, observed public posts, explicit preferences and a narrowly supported meeting register. It produced a detailed private guide and a separate shareable prompt. The reusable method is the skill here. The person's private correspondence stays private.
Read the full article on soxoa.com
Use build-voice-guide on these authorized samples. Build separate cards for operational email and long-form articles. Distinguish authored, edited and synthetic material. Keep all private excerpts out of the shareable prompt.
/plugin marketplace add johnarndt/soxoa-claude-plugins /plugin install soxoa-ops@soxoa-pluginsDownload build-voice-guide.zipView on GitHub
View SKILL.md
--- name: build-voice-guide description: Build a reusable writing voice guide from authorized samples, with channel-specific rules, provenance and a public-safe prompt. --- # Build a voice guide Turn a person's authorized writing samples into a useful guide for future drafting. Describe patterns supported by the corpus and preferences supplied by the person. Do not manufacture a persona from a few messages. ## Establish the corpus Confirm whose voice is being modeled, the intended channels, and which supplied or connected sources the user has authorized. Use available files and connectors; access to one source does not authorize searching every account. Do useful analysis on available material and identify gaps that affect the requested output. Keep a private source ledger recording source ID, date, channel, audience, authorship status, editing/assistance status and permission to quote. Distinguish: - verified person-authored material; - observed publication under that person's account; - attributed, edited or coauthored work; - explicit preferences and corrections; - synthetic drafts and templates. Do not treat the last two categories as proof of natural writing habits. Being sent from the person's account does not establish unassisted authorship. Remove quoted replies, forwarded material, signatures, automated content and unrelated third-party speech before analysis. Deduplicate by stable message ID when available, otherwise by normalized content and context. Preserve enough source references to inspect each conclusion privately. ## Find patterns by register Group samples by channel, audience and stakes before identifying common features. Compare openings, sentence structure, level of explanation, vocabulary, warmth, uncertainty, humor and requested next actions. A terse internal chat should not determine the person's public article voice. For each proposed rule, retain a private supporting reference and a confidence label. Use quantitative summaries only when the cleaned sample and denominator are defined. Mark sparse channels as limited evidence. When preferences contradict historical frequency, explain the difference and follow the person's current preference for future drafting. For transcripts, analyze only attributed language. A transcript alone cannot establish accent, pitch, volume or physical delivery. Do not insert fillers, typos, profanity or disfluencies to simulate likeness. ## Deliver usable instructions Produce a concise core voice description plus register cards relevant to the requested use. Each card should explain the opening, development, detail level, tone and close. Include factual boundaries, an anti-caricature note and a self-contained drafting prompt. Keep private excerpts, personal identifiers and source-analysis details out of the shareable prompt. Provide original synthetic examples labeled as such when examples clarify a rule; do not present them as the person's words. Where the corpus permits, reserve samples from initial analysis and check whether the guide describes them. Report this as a qualitative consistency check, not a validated score for authorship or human likeness. Invite the person to compare a small set of representative drafts; revise recurring mismatches instead of accumulating one-off rules. Return the guide, portable prompt, source-coverage summary and material uncertainty. Creating a guide does not authorize publishing samples or posting in the person's name.