cadence

Solid

Use when the user wants to write, draft, rewrite, edit, or improve prose that sounds human instead of AI-generated — essays, posts, landing copy, emails, docs, scripts, newsletters, threads. Learns a reusable "voice" from a sample the user provides (a book, an article, their own past writing) and writes in it. Strips AI tone: hollow-confidence words, uniform sentence rhythm, triadic lists, negation-pivot clichés, throat-clearing openers. Use for "make this sound less like AI," "write this in the voice of X," "match my tone," "de-slop this draft," or building a library of tones to choose from. Not for code, data, or visual/UI design.

Data & Documents 10 stars 5 forks Updated 2 days ago MIT

Install

View on GitHub

Quality Score: 82/100

Stars 20%
35
Recency 20%
100
Frontmatter 20%
70
Documentation 15%
100
Issue Health 10%
80
License 10%
100
Description 5%
100

Skill Content

# Cadence Cadence makes AI-written prose sound like it was written by a person — and, when the user asks, by a *particular* person or in a *particular* tone they've chosen. Two engines do the work: - **Voice profiles** (the positive target) — a structured spec of how one voice actually moves: its rhythm, diction, devices, stance, and bright lines. Learned from a sample, saved as a portable file, reused on demand. See `reference/voice-profile-schema.md`. - **The de-slop detector** (the negative target) — `scripts/deslop.mjs`, a deterministic analyzer that scores text and names every AI tell in it. It is real code, not a prompt: the same draft scores the same every time, so you can show a before/after. ## Shared writing laws These hold for every command. They are the difference between prose and slop. 1. **Rhythm is the strongest tell.** AI writing clusters every sentence around the same length. Human writing swings — a long accumulating sentence, then a short one. Vary it on purpose. Target a sentence-length CV ≥ 0.5. 2. **Concrete before abstract.** Show the case, then name the principle. A reader should be able to picture something before you ask them to think about it. 3. **No hollow confidence.** Never assert quality with adjectives ("seamless," "robust," "powerful," "comprehensive"). Show the thing; let the reader conclude it's good. 4. **Cut the throat-clear.** No "In today's world," no "It's important to note," no "When it comes to." ...

Details

Author
wuisabel-gif
Repository
wuisabel-gif/Cadence
Created
1 months ago
Last Updated
2 days ago
Language
JavaScript
License
MIT

Bundled in these plugins

Similar Skills

Semantically similar based on skill content — not just same category

Code & Development Listed

human-voice

Edit or write external-facing prose so it doesn't read as AI-generated. Apply to documentation, blog posts, tutorials, public READMEs, customer-facing slide decks, marketing copy, and competitive briefs — whenever the user asks to write, edit, review, or style-check such content, even if they don't mention AI. Apply when creating this kind of content from scratch. Do NOT apply to internal communications (Slack, Notion, internal PRs, meeting notes) unless the user explicitly asks for a style review. Trigger phrases include "write a blog post", "edit this doc", "review this tutorial", "write copy for", "external-facing", "customer-facing", "public README", "slide deck for customers", "competitive brief", "does this sound AI-generated", "check this for AI tells", "make this less AI", "remove the slop".

1 Updated today
reidworks-io
AI & Automation Solid

stop-slop

Load when the user asks to draft, edit, or review English prose for AI-writing tells; do not load for code explanation, Chinese output, technical specs, or routine status reports.

84 Updated 5 days ago
JasonxzWen
Data & Documents Listed

deslop

Strip AI slop out of text while preserving the writer's own voice, or audit text for AI tells without rewriting it. Use whenever the user asks to deslop text, make writing sound less AI-generated or more natural, remove "AI slop", check whether something reads as AI-written, or clean up a draft for sending or publishing (emails, job applications, posts, articles, profiles). Also trigger on mentions of em dash overuse, words like "delve"/"leverage"/"seamless", or requests to make text "sound human". Covers 37 documented patterns from Wikipedia's "Signs of AI writing" plus petergyang/no-ai-slop, and enforces editing fundamentals: minimum effective edit, front-load where it helps, protect specific facts, preserve the writer's opinions and cadence, add voice only where there is none.

1 Updated 4 days ago
devv-shayan