Skip to content

writing.md

Latest commit

 

History

History
47 lines (30 loc) · 2.42 KB

writing.md

File metadata and controls

47 lines (30 loc) · 2.42 KB
Raw

Writing bundle

ai-slop-killer

The final anti-AI-slop editing pass for any prose — blog posts, newsletters, social, email, marketing/web copy, docs, essays, fiction. It's the last thing to touch a draft before it ships.

What it does

Runs a four-step loop: Detect → Score → Fix → Prove.

  1. Detect — scans for documented AI tells: negative parallelism ("it's not X, it's Y"), rhetorical Q&A, rule-of-three and anaphora pile-ups, em-dash and unicode overuse, corporate-AI vocabulary, wind-up intros and signposted conclusions, bold-first-bullet listicles, low-burstiness rhythm, hedging, vague attributions, grandiose stakes, and more.
  2. Score — a 0–100 slop score (lower is cleaner), so the improvement is visible.
  3. Fix — rewrites the tells in place while preserving meaning and voice.
  4. Prove — returns the cleaned text plus an itemized list of what changed and what was left in on purpose.

The catalog is built from documented sources — Wikipedia's Signs of AI writing, the tropes.fyi directory, and working editors' guides — not one person's taste.

Use it

  • "Does this sound like AI? Clean it up."
  • "De-slop this newsletter draft."
  • "Copyedit this and show me what you cut."

Inputs / outputs

  • In: any passage of prose (paste it, or point to a file).
  • Out: the cleaned passage, a before→after score, and the change log.

Configure

Drop a slop.config.md in your working directory to add house rules. Copy the annotated example from plugins/ai-slop-killer/skills/ai-slop-killer/reference/slop.config.example.md. You can set:

  • banned_words — extra words to hard-ban.
  • allowed_words — documented-tell words you legitimately use (kills false positives; every domain has a "robust").
  • voice — a voice/style skill or a short note to preserve.
  • citations — required | preferred | off.
  • channel_defaults — per-channel emoji/format rules.

Reference files

  • reference/slop-catalog.md — the full field guide of tells, with fixes and sources.
  • reference/scoring.md — how the slop score is computed.
  • reference/slop.config.example.md — annotated config template.

Notes

It won't over-correct: one em-dash, one triad, one rhetorical question, one metaphor per piece is human. The goal is prose that reads like a person wrote it on purpose — lived-in, not scrubbed flavorless. It has an optional fiction mode for manuscript prose.