stop-slop

---
name: stop-slop
description: Use when drafting, editing, or reviewing prose so the agent removes predictable AI writing tells and returns direct, specific, human-sounding copy.
source: https://github.com/hardikpandya/stop-slop/blob/main/SKILL.md
generated: 2026-06-29T17:52:26.870Z
category: concept
audience: marketing
---

## When to use

- Cleaning up marketing copy, launch posts, landing-page sections, newsletters, or social drafts that sound AI-written
- Reviewing executive updates, internal docs, or productivity writing for filler, passive voice, and formulaic structure
- Turning generic AI prose into direct prose with concrete actors, specifics, and varied sentence rhythm
- Scoring a draft before publishing using directness, rhythm, trust, authenticity, and density

## Key concepts

### Filler phrase removal

Cut throat-clearing openers, emphasis crutches, business jargon, adverbs, self-referential asides, and vague importance claims. The skill points to a reference list of common offenders.

### Formulaic structure cleanup

Avoid AI-shaped patterns such as binary contrasts, negative listings, dramatic fragments, rhetorical setups, false transformation arcs, and three-item rhythm.

### Human subject discipline

Every sentence should name a human actor when possible. Passive constructions and inanimate abstractions doing human verbs hide responsibility and flatten the prose.

### Specificity over vague authority

Replace sweeping claims, lazy extremes, and sentences like 'the implications are significant' with the exact person, action, constraint, number, or consequence.

### Reader-in-the-room voice

Write to the reader directly instead of narrating from above. 'You' and concrete scenes beat distant generalizations about people, teams, or markets.

### Five-dimension scoring

Before delivery, rate directness, rhythm, trust, authenticity, and density from 1 to 10. Below 35 out of 50 means revise before handing over the draft.

## API reference

```
Install stop-slop from GitHub
```

Use the upstream skill by copying its SKILL.md into the agent's skill directory.

```
git clone https://github.com/hardikpandya/stop-slop
mkdir -p ~/.codex/skills/stop-slop
cp stop-slop/SKILL.md ~/.codex/skills/stop-slop/SKILL.md
```

```
Stop Slop revision pass
```

A reusable prompt shape for applying the skill to any draft.

```
Apply stop-slop to this draft. Remove filler phrases, binary contrasts, passive voice, vague declaratives, em dashes, and pull-quote endings. Keep the meaning. Return the revised draft, then score directness, rhythm, trust, authenticity, and density out of 10.

DRAFT:
<paste text>
```

```
Quick checks before delivery
```

Run a compact checklist after writing or editing prose.

```
Check this draft for adverbs, passive voice, inanimate subjects doing human actions, Wh- sentence starters, 'here is what' openers, 'not X but Y' contrasts, repeated sentence length, punchy one-liner paragraph endings, em dashes, vague significance claims, distant narrator voice, and meta-joiners.
```

## Gotchas

- Do not preserve AI-sounding phrasing just because it is grammatically correct; the point is to remove recognizable patterns.
- Do not remove specificity while compressing. A shorter generic sentence is still slop.
- The skill forbids em dashes, but one upstream example still contains one; follow the rule, not that example.
- Active voice does not mean aggressive voice. Name the actor and action without manufacturing hype.
- A pull-quote ending can sound polished and still feel fake. Rewrite it as a normal sentence or cut it.
- Use the scoring rubric as a revision gate, not as decoration after the draft is already done.

---
Generated by SkillMake from https://github.com/hardikpandya/stop-slop/blob/main/SKILL.md on 2026-06-29T17:52:26.870Z.
Verify against source before relying on details.

File: ~/.claude/skills/stop-slop/SKILL.md