humanizer-prolisted
Install: claude install-skill jodyjdc/ultimate-best-seller-studio
# humanizer-pro (bridge)
This skill does **not** reimplement humanization. It points at the **Better
Humanizer** — a measured, self-improving rewriter that is maintained as its own
project so it never has to be kept in sync in two places.
- **Upstream (single source of truth):** <https://github.com/jodyjdc/better-humanizer>
- **In this repo:** the `external/better-humanizer` git submodule (version-pinned).
Initialize it once with `git submodule update --init --recursive`.
When `install.sh` / `install.ps1` runs, it installs the *full* Better Humanizer
on top of this bridge, so in Claude Code `/humanizer-pro` is the complete tool —
scorer, corpora, judges, registers, and personas — not this pointer.
## Why it replaced the old `humanizer`
The previous `humanizer` skill was a static checklist of ~33 known AI "tells" from
Wikipedia's *Signs of AI writing*. Better Humanizer keeps that pattern knowledge
and adds the three things a checklist cannot do:
1. **Measurement, not vibes.** `scripts/stylo.py` (standard-library, no deps) rates
every rewrite against a real human reference band: sentence-length burstiness,
lexical diversity (MTLD), a function-word fingerprint, punctuation rates, tell
counts, and document-level discourse structure.
2. **Register awareness.** "Human" is register-specific. Seven calibrated registers
ship — for a novel studio the one that matters is **`literary`** (it tolerates
~13× the em dashes of casual prose, expects high burstiness, and