Writing skills monorepo

Three installable skills in one repo:

Skill	Version	Role
humanizer	package 1.3.4	Natural rewrites: cut AI-sounding patterns, keep voice
plain-english	skill 1.0.0	ASD-STE100 simplified technical English + ste_lint.py
ai-writing-detector	skill 1.0.2	Signals-only AI-writing scan + rewrite preservation check

Pick by job:

• humanizer for essays, social, memos, and voice-preserving cleanup
• plain-english for docs, PR text, errors, runbooks, and form-slop cleanup
• ai-writing-detector for scores and issue lists, or to check that an edit left code/URLs/structure alone. Not authorship proof. Not detector evasion.

Humanizer

Turns stiff or generated-sounding prose into clear, specific writing that keeps the author's meaning and voice.

Supports light cleanup, full rewrites, voice-sensitive edits, tone shifts, fiction and dialogue work, detect-only audits, edit-in-place cleanup, and deeper editorial diagnosis. When you need more depth: a core 55-pattern catalog, extended publication tells (through pattern 73), Tier 1A/1B vocabulary tables, never-inject rewrite guardrails, and silent post-edit checks.

What it does

• strips inflated, generic, promotional, or robotic phrasing
• keeps facts, uncertainty, terms of art, and deliberate style
• fits audience, platform, and tone
• works across vocabulary, syntax, paragraphs, point of view, dialogue, setting, and delivery as one system
• runs detect-only audits with severity triage, or minimal in-place file edits
• uses tiered word tables without treating hits as proof of authorship
• silent required-checks QA; optional long-form diagnostics; provenance note
• critiques a draft without rewriting it

Goal: better writing. Not AI-detector evasion. Not fake typos or forced imperfection.

Plain English

ASD-STE100-style controlled writing: short sentences, active voice, plain verbs, no marketing fluff, no em dash. Ships a deterministic linter (scripts/ste_lint.py) that scores violations per 100 words.

Modes: STE-flavored (default for docs/PRs) and strict (procedures, safety, errors).

AI writing detector

Companion to humanizer. Vendors a zero-dependency JS engine, then wraps output in a signals_only report so agents do not read scores as authorship.

node skills/ai-writing-detector/scripts/analyze.js path/to/file.md
node skills/ai-writing-detector/scripts/validate-cli.js before.md after.md
bash skills/ai-writing-detector/scripts/smoke.sh

Default analyze output drops document_classification. For quality rewrites, use humanizer. Do not chase the score.

Repository layout

humanizer/   # git repo name
├── AGENTS.md
├── CHANGELOG.md
├── LICENSE
├── README.md
├── SOURCES.md
└── skills/
    ├── humanizer/
    ├── plain-english/
    └── ai-writing-detector/
        ├── SKILL.md
        ├── agents/openai.yaml
        ├── references/…
        └── scripts/   # patterns.js, validate.js, analyze.js

Each folder under skills/ is a full installable package. Root docs describe the monorepo. Machine-local clone pins live only in SOURCES.md, not inside skill runtime docs.

Installation

Single skill

cp -R skills/humanizer ~/.claude/skills/
# or: ~/.openclaw/skills/  ~/.codex/skills/  etc.

cp -R skills/plain-english DESTINATION/
cp -R skills/ai-writing-detector DESTINATION/

Hermes-style category folder

mkdir -p ~/.hermes/skills/writing
cp -R skills/humanizer ~/.hermes/skills/writing/
cp -R skills/plain-english ~/.hermes/skills/writing/
cp -R skills/ai-writing-detector ~/.hermes/skills/writing/

skills.sh CLI

npx skills add apoapostolov/humanizer --skill humanizer
npx skills add apoapostolov/humanizer --skill plain-english
npx skills add apoapostolov/humanizer --skill ai-writing-detector

Which skill?

Need	Skill
Natural voice, social, essay, fiction polish	humanizer
Docs, PR body, errors, runbooks, form-slop cleanup	plain-english
Score / issue list / keep code+URLs intact after edit	ai-writing-detector
Authorship or integrity claim	provenance evidence, not a detector score

Version

Component	Version
Monorepo / release tag	1.3.4
humanizer skill package	ships with monorepo 1.3.4
plain-english skill	1.0.0
ai-writing-detector skill	1.0.2

Upstream ingest pins: SOURCES.md. Ingest-only humanizer updates bump patch only. New monorepo packaging (new skill folder) bumps minor.

See CHANGELOG.md.

License

MIT. See LICENSE.

Plain-english linter and episode kit credit: skills/plain-english/references/source-and-limits.md. ASD-STE100 stays copyright of its publisher; this repo does not ship the full standard.

Detector engine credit: skills/ai-writing-detector/references/ATTRIBUTION.md.