Skill · Verify
Spot bias and weak arguments
Have an AI audit an article, pitch or argument for logical gaps, loaded language, cherry-picked evidence and missing counterarguments.
The skill#
Act as a critical-thinking instructor reviewing an argument.
The text: [PASTE THE ARGUMENT OR ARTICLE]
1. State the main claim in one sentence.
2. List the reasons and evidence offered for it.
3. Identify logical gaps, unsupported leaps and fallacies, naming each one.
4. Flag loaded or emotional language and what neutral wording would be.
5. Note evidence that seems cherry-picked, and what is missing.
6. Give the strongest counterargument the text ignores.
Judge the reasoning, not whether you agree with the conclusion.
Return:
Main claim
Strengths
Weaknesses (issue | where | why it matters)
Strongest counterargument
Overall: how well the evidence supports the claim (strong, moderate, weak)How to use it#
- Copy the skill into any AI model.
- Paste the article, pitch or post you want to test.
- Use the weaknesses list to decide what to question, check or rewrite.
Example#
Paste an opinion column on remote work. You get the main claim, three named fallacies with the sentences they appear in, two pieces of anecdotal evidence presented as general proof, and the strongest counterargument the author skipped.
Quick and strict versions#
Use the quick version for everyday checks and the strict version when the result matters.
In five bullet points, list the weakest parts of this argument: gaps, loaded words or missing evidence. Judge the reasoning, not the conclusion.
Text: [PASTE THE TEXT]Act as a debate judge and rhetoric expert. Text: [PASTE THE TEXT]
1. Map the argument: claim, premises, evidence, conclusion.
2. Test each premise: is it stated, supported, or assumed?
3. Name every fallacy with the exact sentence where it occurs.
4. Rate each piece of evidence for relevance, reliability and representativeness.
5. Identify framing choices: what is emphasized, omitted or assumed about the reader.
6. Steelman the opposing view in its strongest form.
Stay neutral on the conclusion. Quote the text for every criticism.
Return: Argument map | Issue table (quote | problem | severity) | Steelman | VerdictUse it as an agent skill#
Save this as SKILL.md in a folder named after the skill, and agents that support skills (Claude, Claude Code and other skill-aware agents) can load it automatically when the task fits.
---
name: spot-weak-arguments
description: Use when the user wants an argument, article, pitch or opinion piece checked for bias, fallacies or weak reasoning.
---
# Spot bias and weak arguments
Fill each [BRACKETED] field from the conversation. If one is missing, ask the user for it before starting.
Act as a critical-thinking instructor reviewing an argument.
The text: [PASTE THE ARGUMENT OR ARTICLE]
1. State the main claim in one sentence.
2. List the reasons and evidence offered for it.
3. Identify logical gaps, unsupported leaps and fallacies, naming each one.
4. Flag loaded or emotional language and what neutral wording would be.
5. Note evidence that seems cherry-picked, and what is missing.
6. Give the strongest counterargument the text ignores.
Judge the reasoning, not whether you agree with the conclusion.
Return:
Main claim
Strengths
Weaknesses (issue | where | why it matters)
Strongest counterargument
Overall: how well the evidence supports the claim (strong, moderate, weak)
Frequently asked questions
Will the AI just disagree with the author?
The skill tells it to judge the reasoning, not the conclusion, and to name strengths too. Check that its criticisms quote the text.
Can I use it on my own writing?
Yes, and it is one of the best uses: run it before you publish to find the gaps a critic would.
Does AI have its own bias?
It can. Asking it to steelman the other side and to quote the text for every point reduces, but does not remove, that risk.