Skip to content

Releases: solanabr/writer-style-skill

Release list

v1.1.0 — Empirically calibrated humanization

v1.1.0 — Empirically calibrated humanization Latest
Latest

Choose a tag to compare

View all tags
@kauenet kauenet released this 04 Jul 16:45
v1.1.0
384ce0a

v1.1.0 — 2026-07-04

Empirically calibrated humanization (4 blind-judged rounds, 100+ adversarial agents):

  • Seams≠markers architecture: markers: budgets + context gates in the card; kaue.md §5b identity anti-triggers (biography exempt, named receipts win)
  • themes.md substance bank (receipted stances/anecdotes; no hyperspecific third-party names)
  • profiles/kaue/LESSONS.md — calibrated writing lessons, writer-loaded and binding for ALL routed blends (ROUTING guard + per-secondary guards)
  • Pipeline v2: mood + re-outline-in-voice, coordinator palettes, fresh-eyes reviser, long-form ledger, course mode (context-gated closers/openers)
  • Validator v2: density (marker budgets/gates), intra-doc audit, warmth telemetry, batch twin-scan, QA-machine/epigram/clipped-sentence checks
  • Golden regression outputs in testbed/accepted/ (check_round.py)
  • Calibration working data moved out of the repo (gitignored); knowledge absorbed into pack/rules
Assets 2
Loading

writer-style-skill v1.0.0

writer-style-skill v1.0.0

Choose a tag to compare

View all tags
@kauenet kauenet released this 27 Jun 02:56
v1.0.0
37f4941

First public release.

A Claude Code skill that writes original educational and long-form technical content in a specific author's voice — facts verified first, AI tells engineered out. Ships the kaue pack (Kaue / Superteam Brazil, tuned for Solana/Web3 education) as the default; the engine is voice-agnostic.

Install

Claude Code plugin (recommended — every project):

claude plugin marketplace add solanabr/writer-style-skill
claude plugin install writer-style-skill@stbr

npm (per-project):

npm add @stbr/writer-style-skill

What's inside

  • Self-contained skill at skills/writer-style/ — a primary voice (Kaue, 14-exemplar bank, always on) plus 5 secondary craft voices (helius, vitalik, balaji, hayes, hotz) routed by the piece's job; idiolect reproduced for the primary, stripped for the secondaries.
  • Facts first, voice last — a 3-pass workflow that verifies facts in neutral prose before styling, so the voice can never alter a number, code line, or API.
  • Pure-Python Pass-C validator (no dependencies): a fact-preservation diff, an AI-tell + tiered-deslop lint, a sentence-cadence floor, and a cross-lesson repetition audit. A portable $SKILL resolver runs it correctly from a plugin, an npm project install, or a clone.

MIT — Superteam Brazil / Kaue.

Loading