Writer-Style

Vendored from solanabr/writer-style-skill (MIT — Superteam Brazil / Kaue).
See LICENSE in this directory and pack ATTRIBUTION.md.
Route via the writing hub when unclear; use writing-prose for non-persona house tone.

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.

Default pack in this install: evan (profiles/evan/) — Evan Kim, calibrated from evan_writings. Upstream also ships kaue (profiles/kaue/) for Superteam Brazil / Solana education — use when the user asks for Kaue or Superteam voice.

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 (this install)

Everything below is the engine; the pack supplies the actual voices. Examples in this file that mention kaue paths work the same with evan — swap the profile directory.

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

• Primary voice (profiles/<pack>/<name>.md + card), 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. Default primary here: profiles/evan/evan.md + evan.card.yaml.
• Secondary voices (craft under profiles/kaue/secondary/ — shared craft lanes; idiolect stripped). Applied in the primary’s voice, never impersonation. One leads (backbone), at most one guests. Chosen by the piece's job, not its topic. Evan pack routing: profiles/evan/ROUTING.md.

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:

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

• evan pack: profiles/evan/ROUTING.md
• kaue pack: profiles/kaue/ROUTING.md

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

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)

Default (evan): profiles/evan/evan.md · evan.card.yaml · modes.md (pick mode by job) · themes.md · mode-matched exemplars from exemplars/evan/ per profiles/evan/ROUTING.md · optional secondary from profiles/kaue/secondary/.

Kaue (when requested): profiles/kaue/kaue.md · kaue.card.yaml · themes.md · exemplars under exemplars/kaue/ · secondary per profiles/kaue/ROUTING.md.

Exemplars are rhythm donors, never content donors. 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/evan/evan.card.yaml" (or the active pack’s card). 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/evan/evan.card.yaml" --facts <fact-sheet> (swap card path for other packs). 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 the loaded pack’s spine + exemplars (evan.md by default).

Progressive disclosure (read when needed)

Agents & commands