Stop Slop: AI Writing Pattern Eliminator
A rule set that strips predictable AI tells out of any draft, from throat-clearing openers to false-agency verbs and em dashes
AI prose has a tell, and readers catch it fast. The em dashes, the 'not X, it's Y' contrasts, the fragments dropped for fake drama. This playbook hands Claude eight rules and a checklist so your writing sounds like you wrote it, not a model.
Who it's for: bloggers editing AI drafts, founders writing essays and updates, ghostwriters, newsletter authors, content marketers, anyone shipping AI-assisted prose
Example
"Rewrite this essay to sound human" → A revision with adverbs cut, passive voice made active, em dashes removed, false-agency verbs assigned to real people, plus a 1-50 score across directness, rhythm, trust, authenticity, and density
CLAUDE.md Template
New here? 3-minute setup guide → | Already set up? Copy the template below.
npx ccplaybooks add stop-slop# Stop Slop
Eliminate predictable AI writing patterns from prose. Use these rules when drafting, editing, or reviewing any text so it reads like a person wrote it, not a model.
## Core Rules
1. **Cut filler phrases.** Remove throat-clearing openers, emphasis crutches, and all adverbs. See the phrase list below.
2. **Break formulaic structures.** Avoid binary contrasts, negative listings, dramatic fragmentation, rhetorical setups, false agency. See the structure list below.
3. **Use active voice.** Every sentence needs a human subject doing something. No passive constructions. No inanimate objects performing human actions ("the complaint becomes a fix").
Get new playbooks like this one
One email a week with new Claude Code workflows. Free, like everything here.
No spam. Unsubscribe anytime.
What This Does
Gives Claude a concrete rule set for catching and removing the patterns that make prose sound machine-generated. It works on three levels: banned phrases (throat-clearing openers, business jargon, adverbs), structural clichés (binary contrasts, negative listing, false agency), and sentence-level rules (no Wh- starters, no em dashes, active voice only). After a rewrite, Claude scores the result 1-10 across five dimensions so you know whether it still needs work.
The rules come from Hardik Pandya's "stop-slop" skill. This packages them as a CLAUDE.md so any folder becomes a slop-detecting editor.
Quick Start
Step 1: Create a Project Folder
mkdir -p ~/Documents/StopSlop
Step 2: Download the Template
Click Download above, then:
mv ~/Downloads/CLAUDE.md ~/Documents/StopSlop/
Step 3: Start Working
cd ~/Documents/StopSlop
claude
Paste a draft and ask Claude to rewrite it, or drop a .md file into the folder and say "clean the slop out of this."
The Eight Core Rules
- Cut filler phrases — throat-clearing openers, emphasis crutches, and every adverb.
- Break formulaic structures — no binary contrasts, negative listings, or dramatic fragmentation.
- Use active voice — every sentence gets a human subject doing something.
- Be specific — name the thing instead of announcing that it matters.
- Put the reader in the room — "you" beats "people," specifics beat abstractions.
- Vary rhythm — mix sentence lengths, two items beat three, no em dashes.
- Trust readers — state facts directly, skip the softening.
- Cut quotables — if it sounds like a pull-quote, rewrite it.
The Scoring Rubric
After every rewrite, Claude rates the draft 1-10 on five dimensions and adds them up:
| Dimension | Question it answers |
|---|---|
| Directness | Statements or announcements? |
| Rhythm | Varied or metronomic? |
| Trust | Respects reader intelligence? |
| Authenticity | Sounds human? |
| Density | Anything cuttable? |
A total below 35/50 means the draft still needs a pass. The score gives you a fast signal without re-reading the whole thing.
What It Catches
The template ships three reference sets so Claude flags patterns by category:
- Phrases — "Here's the thing," "It turns out," "the uncomfortable truth is," plus jargon like "lean into," "deep dive," and "circle back," and adverbs like "really," "just," and "genuinely."
- Structures — binary contrasts ("not X, it's Y"), negative listing ("Not a X... Not a Y... A Z"), and false agency ("the complaint becomes a fix," "the decision emerges") where an inanimate thing does a human's job.
- Examples — five before/after rewrites that show the rules applied, so Claude has a model to match rather than abstract instructions.
Examples
Throat-clearing plus binary contrast:
Before — "Here's the thing: building products is hard. Not because the technology is complex. Because people are complex. Let that sink in."
After — "Building products is hard. Technology is manageable. People aren't."
Business jargon stack:
Before — "In today's fast-paced landscape, we need to lean into discomfort and navigate uncertainty with clarity."
After — "Move faster. Your competition is."
Rhetorical setup:
Before — "What if I told you that the best teams don't optimize for productivity? Here's what I mean: they optimize for learning."
After — "The best teams optimize for learning, not productivity."
Tips
- Use it in reverse to detect AI writing: ask Claude to flag which rules a draft breaks and where, without changing the text.
- The false-agency rule is the highest-value one. Naming a real person behind every verb fixes more slop than swapping vocabulary.
- Run the score twice, before and after, to see the delta. A jump from 22 to 40 is more convincing than "looks better."
- Keep your own voice notes in the same folder so Claude edits toward how you actually write, not a generic clean style.
Limitations
- The rules are absolutes by design (no em dashes, no adverbs, no three-item lists). Loosen them for cases where the flagged form is the right call.
- Aggressive cutting can strip nuance from technical or legal writing. Review edits before shipping anything precise.
- It targets English prose patterns. Tells differ in other languages.
- A high score means the text avoids known AI patterns, not that the argument is good or the facts are correct.