Stop Slop English writing skill

September 15, 2026 · 3 min read

If you have any questions, feel free to comment below. Click the block can copy the code.
And if you think it's helpful to you, just click on the ads which can support this site. Thanks!

Stop Slop^[1] is a lightweight, open-source prompt skill for editing English prose that exhibits predictable LLM writing patterns. It provides Markdown instructions and examples for making generated or revised text more direct, concrete, and editorially restrained.

Background and Goals #

Stop Slop targets recurring patterns in LLM-generated prose: stock phrases, exaggerated emphasis, empty abstraction, formulaic rhetorical turns, passive constructions, and repetitive sentence cadence. The project guides a model to identify and remove those patterns.

Core Features #

Technical Architecture #

Its architecture is a prompt-resource model:

Writer, editor, or software workflow
                |
                v
LLM host or agent runtime
(Claude Code, Claude Project, API, other LLM)
                |
                v
SKILL.md
  |- Core editing policy
  |- Quick quality checks
  |- 50-point rubric
                |
                +--> phrases.md
                +--> structures.md
                +--> examples.md
                |
                v
Draft, review comments, or revised prose

The project uses Markdown for instructions and YAML front matter for skill metadata.

Example #

I summarized and distilled the following instructions from the repository’s content. In chat-only settings, you do not need to load the skill; paste this text after your writing or editing request.

Paste these instructions after your writing or editing request. They remove common AI writing patterns while preserving meaning, voice, and necessary nuance.

Follow these rules when drafting or revising prose:

Start with the point. Remove throat-clearing phrases such as “Here’s the thing,” “It turns out,” “The truth is,” “Let me be clear,” “This matters because,” and “It’s worth noting.” State the fact or claim directly.

Name the actor. Use active voice when someone makes a decision, changes something, discovers a result, or solves a problem. Replace false agency such as “the decision emerged” with “the team decided.”

State conclusions directly. Avoid formulaic reversals such as “Not X, but Y,” “X isn’t the problem; Y is,” and “What if I told you…?” Present the claim and its evidence without manufacturing a reveal.

Replace vague judgments with specifics. Name the cause, consequence, number, person, or observable behavior. Replace statements such as “The implications are significant” with the actual implication. Use absolute words such as “every,” “always,” and “never” only when they are accurate.

Cut empty language. Remove adverbs, intensifiers, hedges, and business jargon when they add no information. Prefer plain verbs: “handle” over “navigate challenges,” “examine” over “deep dive,” and “next” over “moving forward.”

Vary the rhythm. Mix sentence lengths and paragraph endings. Reduce three-part lists, stacked fragments, questions answered immediately, and paragraphs that end with manufactured one-line revelations. Use commas or periods for natural pauses.

Trust the reader. Show the facts and actions, then let the reader draw conclusions. Remove meta-commentary, structural announcements, unnecessary reassurance, repeated explanations, and forced takeaways.

Preserve the source material’s facts, position, voice, and necessary qualifications. Apply each rule according to context. Clarity and accuracy take priority over mechanical compliance.

Example:

Before:

Here’s the thing: building products is hard. Not because the technology is complex, but because people are complex. Let that sink in.

After:

Building products is hard. Technology can be tested. Customers change their minds, and teams disagree about what to build.

Before:

The decision emerged after the data revealed a significant shift in the market landscape.

After:

The product team reviewed the sales data and cancelled the expansion.

Before delivering the result, check:

Does the opening begin with substance?

Does each important action have a clear actor?

Have vague judgments been replaced with specific facts?

Does every sentence add information?

Do sentence and paragraph rhythms vary naturally?

Does the piece end where the content ends?

Return only the revised prose unless the user asks for an explanation.