# Skill: Run the slop detection checklist on any text in under two minutes

> Lets an AI agent apply a five-marker diagnostic plus a replaceability test to any piece of content,
> return a per-marker verdict, and provide actionable fixes; applies when evaluating content before
> publication or assessing a contractor's or competitor's work.

Source: sodigi·learn - neuroslop/kak-raspoznat-slop · https://sodigi.io/learn/neuroslop/kak-raspoznat-slop

## When to use
- Before the user publishes any text (post, article, email, landing page, case study).
- When evaluating a contractor's draft or case studies to decide whether to hire them.
- When sizing up a competitor's or "expert's" content for quality.
- When a user asks "is this text good?" and needs a specific, actionable answer rather than a vague impression.

## Core rules
- The master test is replaceability: if the product/person/company can be swapped for any other and the text still holds, it says nothing specific about anything.
- One marker failing is a coincidence. Two is a warning. Three or more is a confirmed slop diagnosis.
- Quality content does not look like "AI content" - it looks like the work of someone who knows their subject, with numbers and a position.
- Slop is not the same as brevity or simplicity. A three-sentence post with one precise fact and a clear claim is not slop.
- Structure (subheadings, bullets, tables) is not evidence of substance. Test what is inside the structure, not the structure itself.
- The checklist applies to visual content by the same logic: a generic illustration that fits any brand has failed the replaceability test.
- The checklist tests the result, not the production process. AI involvement is irrelevant to the verdict.

## Procedure
1. Apply the replaceability test first (10 seconds): mentally swap the named product, person, or company for a different one. If the text reads unchanged, the text fails.
2. Check Marker 1 - Specifics: scan for numbers, named companies, named people, dates, verifiable examples. No facts that can be checked = FAIL.
3. Check Marker 2 - Position: find one sentence someone could argue against. "Remote work has pros and cons" is not a position. "For teams under ten, an office is wasted money" is. No arguable claim = FAIL.
4. Check Marker 3 - Cliches: read aloud the first and last paragraphs. Flag every worn phrase ("in today's world", "it's no secret", "real game-changer", "level up"). Entire text assembled from them = FAIL.
5. Check Marker 4 - Structure for show: strip all formatting mentally. Count whether meaningful content remains. Subheadings with three filler lines each = FAIL.
6. Check Marker 5 - Stake: look for any moment the author could be proven wrong: a prediction, an admitted mistake, a named position. No such moment = FAIL.
7. Tally: 0-1 FAILs = not slop; 2 = borderline, flag for improvement; 3+ = slop diagnosis, return specific fixes for each failing marker.

## Ready-to-use prompt
```
You are a content quality auditor. Run the following 6-point slop check on the text below and return a structured report.

REPLACEABILITY TEST
Swap the main subject (product / person / company) for a different one. Does the text still read as correct? Answer YES (fail) or NO (pass) with one sentence of explanation.

MARKER 1 - SPECIFICS
List every verifiable fact present (numbers, names, dates, real examples). Verdict: PASS or FAIL.

MARKER 2 - POSITION
Quote the strongest arguable claim in the text. If none exists, state "none found". Verdict: PASS or FAIL.

MARKER 3 - CLICHES
List every cliche or stock phrase found. Verdict: PASS (0-2 cliches) / WARN (3-4) / FAIL (5+).

MARKER 4 - STRUCTURE VS SUBSTANCE
Describe what content remains if all formatting (headings, bullets, bold) is removed. Verdict: PASS or FAIL.

MARKER 5 - STAKE
Identify any moment the author could be proven wrong. If none, state "none found". Verdict: PASS or FAIL.

FINAL VERDICT: NOT SLOP / BORDERLINE / SLOP
For every FAIL: give one concrete fix the author can apply in under five minutes.

Text:
[PASTE TEXT HERE]
```

## Pitfalls
- Do not diagnose slop from a single failing marker alone; require at least three concurrent failures.
- Do not equate "written by AI" with slop; the diagnostic is about content properties, not origin.
- Do not trigger the cliche marker on one or two worn phrases; every text has a few. The marker fires when the entire text is assembled from them.
- Do not let structure deceive you: seven subheadings and a numbered list look authoritative but can mask total emptiness.
- Do not skip the self-check use case: the most valuable application is running the checklist on the user's own draft before publishing, not only on other people's content.
- Do not recommend adding more words to fix slop; the fix is almost always adding specific facts (numbers, cases, named examples) and cutting filler.