Prompts to "unslop" your content
This resource is based on A copy-paste prompt that edits the slop out of any document, featuring Tim Metz of Animalz, published on the AI Lab by ActiveCampaign.
Get the prompt template
How to use these prompts
These prompts are based on Tim Metz’s “unslop” command—a custom Claude prompt he built at content marketing agency Animalz to cut low-quality filler from AI-generated documents. Copy, paste, and adapt the bracketed variables. The unslop command works on any document type: internal memos, brand kits, strategy docs, email drafts, or content briefs.
Prompt 1: The core unslop command
- Best for: Cleaning up any AI-generated or AI-assisted document
- Use with: Claude (originally built for Claude; works with any LLM)
Unslop command
Review this document and apply three editing principles:
1. MECE (Mutually Exclusive, Collectively Exhaustive): Each section should address a distinct topic with no overlap. Together, all sections should fully cover the subject. Cut or merge any sections that overlap.
2. DRY (Don’t Repeat Yourself): If something has been said once, don’t say it again. Eliminate anything that just rephrases earlier content.
3. As simple as possible, but not simpler: Cut unnecessary complexity and filler, but don’t remove context, examples, or details that are actually needed. If cutting makes the text read like shorthand, you’ve gone too far.
For each change:
- Identify sections to cut, with justification
- Identify sections to relocate, with reasoning
- Then rewrite the document with slop removed
Document to unslop:
[PASTE YOUR DOCUMENT HERE]
Variables to fill in:
- [PASTE YOUR DOCUMENT HERE]—The full text of the document you want to clean up
What to expect: The model will first suggest cuts and relocations with reasoning, then produce a rewritten version. Compare the before and after—if the rewrite reads like shorthand or cuts examples that matter, tell the model to restore specific sections.
Follow-up prompt
The rewrite cut too aggressively in [SECTION NAME]. Restore the examples in that section—they’re important for context. Keep all other changes.
Prompt 2: Slop detector (diagnosis only)
- Best for: Auditing a document before deciding what to change
- Use with: Claude / ChatGPT / Any LLM
Workslop detector
Analyze this document for “workslop”—content that seems valuable at first read but contains no real substance. Flag:
- Sections that overlap or repeat each other (violates MECE/DRY)
- Sentences that sound eloquent but say nothing specific
- Filler transitions that could be cut without losing meaning
- Jargon or buzzwords used in place of concrete details
For each flag, quote the specific text and explain why it’s slop.
Do NOT rewrite anything yet—just diagnose.
Document:
[PASTE YOUR DOCUMENT HERE]
Variables to fill in:
- [PASTE YOUR DOCUMENT HERE]—The document to audit
What to expect: A list of flagged passages with explanations. Use this to decide what to fix manually vs. what to hand back to the model for rewriting.
Prompt 3: Brand kit / system prompt unslopper
- Best for: Cleaning up the brand kits, system prompts, or reference docs that feed into your AI workflows
- Use with: Claude / ChatGPT / Any LLM
Brand kit / system prompt unslopper
This is a [DOCUMENT TYPE] that feeds into an AI workflow. It needs to be clear, non-redundant, and complete — because any slop here compounds in every output the workflow produces.
Apply these principles:
- MECE: Each section covers a distinct topic. No overlaps.
- DRY: Nothing is said twice.
- Essential: Remove filler, but keep every example and specific detail that gives the AI context it needs to produce good output.
After unslopping, list what you changed and why. If you’re unsure whether something should be cut, flag it instead of cutting it.
[PASTE YOUR BRAND KIT / SYSTEM PROMPT / REFERENCE DOC HERE]
Variables to fill in:
- [DOCUMENT TYPE]: e.g., “brand voice guide,” “system prompt,” “editorial style guide,” “LinkedIn content brief”
- [PASTE YOUR BRAND KIT / SYSTEM PROMPT / REFERENCE DOC HERE]: The full reference document
What to expect: A cleaned version plus a changelog. Pay special attention to flagged items—those are the judgment calls where Metz recommends running standardized tests to see if quality improves or regresses after the change.
Prompt 4: Post-unslop quality check
- Best for: Verifying you haven’t cut too much
- Use with: Claude / ChatGPT / Any LLM
Unslop quality check
Compare these two versions of [DOCUMENT TYPE]:
ORIGINAL:
[PASTE ORIGINAL]
UNSLOPPED VERSION:
[PASTE REVISED VERSION]
Check for:
- Does the unslopped version read like shorthand? (If so, it was cut too aggressively)
- Were any important examples removed? (Examples are often mistakenly flagged as unnecessary)
- Does every section in the original still have its core idea represented?
- Is anything important missing that was in the original?
List any problems found. For each, quote what was lost and recommend whether to restore it.
Variables to fill in:
- [DOCUMENT TYPE]: What kind of document this is
- [PASTE ORIGINAL]: The pre-unslop version
- [PASTE REVISED VERSION]: The post-unslop version
What to expect: A gap analysis. Metz notes that when unslopping goes too far, “it starts to read like shorthand” or “starts to cut out examples because it thinks those are unnecessary.” This prompt catches those regressions.
Tips for better results
- Unslop your system prompts first. Metz found that slop in brand kits and reference docs compounds—every output the workflow produces inherits that slop. Clean the inputs before cleaning the outputs.
- Run standardized tests after changes. Metz recommends comparing outputs before and after unslopping a brand kit to confirm quality improved, not regressed.
- Don’t skip the follow-up. The first unslop pass often cuts too aggressively. Use Prompt 4 to catch what was lost, then selectively restore.
- Feed corrections back to the model. Don’t just fix the output. Tell the model what you changed and why. “I wouldn’t say it like that” is better training data than silently rewriting.
Ready for the full story?
Read A copy-paste prompt that edits the slop out of any document, featuring Tim Metz of Animalz, published on the AI Lab by ActiveCampaign.
Get resources to help you replicate Tim’s unslop system:
Related
More data from the AI Lab.