Writer style

Writer-Style

Write original content that sounds like a specific person, not generic AI. And, for technical topics, that is factually correct first and styled second. The engine is voice-agnostic; it ships with the kaue pack (Kaue / Superteam Brazil, tuned for Solana/Web3 education) as the default.

What this skill is for

Use it whenever the output should read like a named author:

• Educational Solana/Web3 content: courses, lessons, modules, explainers, tutorials, deep-dives, threads.
• Long-form technical writing in general, in a target voice.
• Drafting or editing/restyling existing copy into the voice.
• Building a new voice pack (add an author, build someone's voice, design a router) → the persona-builder agent.

Do NOT use for: brand-neutral/house-style copy; content that should sound generic; one-off "summarize this" where voice doesn't matter.

Honest scope. This produces content in the author's register, with AI tells engineered out and facts verified first, not an indistinguishable clone. Voice fidelity is corpus-bounded and improves as more of the author's on-register writing is added. (The default Kaue pack is calibrated from a small on-register sample; lean on the exemplars and the naturalness floor, and don't over-promise "it's exactly him.")

The default pack

Requests like "write this lesson in my voice", "in Kaue's voice", or "a Superteam-style explainer" → use the kaue pack (profiles/kaue/). For another author/domain, load that pack instead; to create one, use the persona-builder. Everything below is the engine; the pack supplies the actual voices.

The two-layer model (summary — full detail in two-layer-model.md)

• Primary voice (profiles/kaue/kaue.md + kaue.card.yaml), the target author's own voice, always on, reproduced (idiolect is the asset). It owns the tone, the seams, and the naturalness floor. Free, not counted toward the layer budget.
• Secondary voices (profiles/kaue/secondary/*.md + cards), transferable craft borrowed from master writers, idiolect stripped (applied in Kaue's voice, never impersonation). One leads (backbone), at most one guests. Chosen by the piece's job, not its topic.

The inversion is the whole trick: reproduce the primary's idiolect, strip the secondaries'. Keep the primary's quirks; borrow only the secondaries' craft.

The workflow — facts first, voice last (full detail in writing-workflow.md)

Persona prompting boosts voice but degrades factual accuracy — fatal for technical content. So voice is the last transform over verified facts, never the medium facts are discovered in. Five steps:

Step	What	Voice
A: Facts	Outline + every claim, number, code line, version-sensitive API, in plain neutral prose. Ground against the Solana doc MCPs (solana-dev, context7, Helius), not model memory. Output: a fact-sheet.	OFF
Gate: Verify	Fact-check the terse fact-sheet before any styling. (Cheap: facts aren't buried in warm prose.) Nothing proceeds until facts are frozen.	—
B: Voice	Load kaue.md + kaue.card.yaml + ~4 exemplars + the routed secondary. Check the card's markers: gates (identity beats need their gate keywords in the fact-sheet — write the marker budget ledger first) and scan themes.md for fitting substance. Restyle the frozen facts: "every number, code line, and named API is frozen; rephrase around them, never change them." Apply the naturalness floor (incl. the plainness quota). Long-form (≥1,200w): section-by-section with the ledger.	ON
B.5: Blind-compare	Re-read the draft against the loaded exemplars (seam + opener) — "him, or a rule-follower?" If it's even/polished/definition-led, regenerate toward a felt-pain opener + uneven rhythm + rotating seams. Run the FIT test on every marker: does it land on a genuine payoff, in this piece's mood, within budget? Keep what fits; never necessity-test personality. Loaded exemplars only, never the corpus (at write-time the corpus is token-waste).	ON
C: Lint	AI-tell scan + fact-preservation diff (no number/identifier mutated A→C) + marker density/gates (an identity beat with no gate keyword in the fact-sheet or a doubled sign-off is a hard fail) + repetition audit (intra-doc for long pieces). A changed 0.002 SOL or renamed instruction is a hard fail.	—

tools/validate_voice.py runs the Pass-C checks (tells, density, diff, audit); the voice-validator agent wraps them.

Running the bundled tools (any install). The validator and data live inside this skill's own directory, so call them with absolute paths. Resolve the directory once and let $SKILL stand for it below ($CLAUDE_PLUGIN_ROOT/skills/writer-style in a plugin, .claude/skills/writer-style in an npm/project install run from the project root, or skills/writer-style in a clone run from the repo root):

SKILL="${CLAUDE_PLUGIN_ROOT:+$CLAUDE_PLUGIN_ROOT/skills/writer-style}"
[ -d "$SKILL" ] || SKILL=".claude/skills/writer-style"; [ -d "$SKILL" ] || SKILL="skills/writer-style"
python3 "$SKILL/tools/validate_voice.py" tells --file <draft> --card "$SKILL/profiles/kaue/kaue.card.yaml"

Routing — which secondary voice (full router in profiles/kaue/ROUTING.md)

Route on the piece's dominant job, not its topic. Each voice is a full backbone in its lane:

The piece's dominant job	Lead voice
Show how a documented Solana system works: tutorial, integration, the runtime, the real artifact	helius
Derive why a deep / non-obvious design is right: protocol internals, cryptoeconomics, mechanism design	vitalik
Frame a broad, interconnected thesis: connect tech/history/economics/society, the big picture & trajectory	balaji
Make macro / market forces legible: economics, tokenomics, geopolitics, incentives, crowd-psychology	hayes
Demystify a hyped / over-complicated thing: collapse it to its true simple model (honest "X is just Y")	hotz
Security / exploit classes	helius (vulnerable→patched) or vitalik (why the check is necessary), never hotz
Pure motivation / "why this matters" / course-opener	primary only + ≤1 light guest

Rules: primary always on & uncounted; cap = backbone (1) + guest (≤1) = 2 secondary layers (3 only for 3,000+ words, segregated; else split the piece); stack only across different lanes; helius is the safe universal guest; hold each voice's guardrail (Hayes builder-optimist not cynical · Balaji peer not oracular · Hotz inviting not dismissive · Vitalik warm not flat). The lanes self-differentiate: no author-vs-author special rules.

The naturalness floor (protect above all craft — rules/naturalness.md)

The body of every craft, applied thoroughly, smooths prose into something that sounds like nobody. Counter it: write unevenly: long calm stretches, enthusiasm only at the edges (open / seams / verdicts / close), a human seam in every passage (a confession, a real number from your own use, a named credit, a "this ran long"). Even enthusiasm + even polish is the #1 AI tell. Always load the seam exemplar.

What you load at generation time (Pass B)

profiles/kaue/kaue.md (primary prose) · kaue.card.yaml (dials + markers: budgets/gates) · profiles/kaue/themes.md (the substance bank — pull FITTING stances/anecdotes instead of recycling exemplar content) · ~4 exemplars from exemplars/kaue/ (always the seam + opener + close + 1 body slot by job; exemplars are rhythm donors, never content donors) · the one routed secondary (secondary/<v>.md + card + its move-demos). Never read evidence/*.profile.json or calibration/ — builder/validator-only.

Self-check before returning (Pass C)

• Burstiness: sentence lengths vary hard (≥1 short punch & ≥1 long run per section); stdev clears the card's burstiness_min. Uniform cadence is the top tell.
• AI tells: run python3 "$SKILL/tools/validate_voice.py" tells --card "$SKILL/profiles/kaue/kaue.card.yaml". It enforces this voice's burstiness floor, em-dash cap, false-antithesis cap, avoid lists, AND a deslop scan (gated cliché tiers with plain-word swaps, crypto-boilerplate, copula/gloss, machine-paste fingerprints, Markdown hygiene).
• Deslop judgment (rules/deslop.md): clear cliché clusters, run the paragraph-reshuffle / "what's-new" / read-aloud tests, but keep the specific detail, mixed feelings, and asides that read human.
• Naturalness: a human seam in every passage; enthusiasm spiked at edges, flat in the body; ≥1 in 3 sections marker-free (the plainness quota).
• Markers: run python3 "$SKILL/tools/validate_voice.py" density --file <draft> --card "$SKILL/profiles/kaue/kaue.card.yaml" --facts <fact-sheet>. An identity beat with no gate keyword in the fact-sheet, or a doubled sign-off, is a hard fail; for ≥1,200w also run audit --file <draft> (section-to-section repetition).
• Facts: every number/identifier from the fact-sheet survived styling unchanged (run python3 "$SKILL/tools/validate_voice.py" diff).
• In-voice: reads like the primary against kaue.md + the loaded exemplars.

Progressive disclosure (read when needed)

Topic	File
The two-layer model, the inversion, the cap, how packs plug in	two-layer-model.md
The facts-first Pass A/Gate/B/C workflow + MCP grounding	writing-workflow.md
The style-card schema, exemplar convention, builder/writer/validator boundary	style-card-schema.md
The full router (per-lane triggers, stacking, failure modes)	profiles/kaue/ROUTING.md
The primary voice (Kaue)	profiles/kaue/kaue.md
The substance bank (stances/anecdotes/analogies with receipts)	profiles/kaue/themes.md
A secondary voice's craft	profiles/kaue/secondary/<voice>.md
Building/refreshing a voice (the adversarial method)	authoring-personas.md → method/
Empirical calibration (the 8-brief testbed + owner-feedback rounds)	testbed/MATRIX.md + profiles/kaue/LESSONS.md (calibration working data is local-only)

Agents & commands

Agent	Purpose	Model
voice-writer	Run the facts-first 3-pass workflow to draft/restyle a piece	opus
persona-builder	Build/refresh a voice pack (adversarial, evidence-bound)	opus
voice-validator	Repetition audit + AI-tell lint + fact-preservation diff	sonnet

Command	Purpose
/write-in-voice	Draft a piece in a pack's voice (facts-first)
/new-persona	Build a new voice via the persona-builder
/profile-corpus	Regenerate builder-internal evidence + card dials
/validate-voice	Run audit / tells / density / diff on generated output
/calibrate-voice	Run a calibration round: regenerate the 8-brief matrix, collect owner ratings, codify changes