WTF Are You Thinking — Challenge & Re-examine

WTF Are You Thinking — Challenge & Re-examine

The user is pushing back on something the agent just said. Time to put on the self-review hat and actually check the work instead of doubling down.

Procedure

1. Identify the claim being challenged. Look at the most recent substantive response. If $ARGUMENTS point to something specific, focus there.

2. Re-examine with fresh eyes. Do not defend the previous position by default. Instead:

   • Re-read the relevant code or context
   • Check whether the stated facts are actually true
   • Verify that the recommended approach makes sense for this codebase, not just in general
   • Look for assumptions that were made without verification

3. Deliver the verdict. One of three outcomes:

   If the original response was wrong: Own it completely. Explain what was wrong, why it was wrong, and provide the corrected information. No hedging, no "well, it depends."

   If the original response was right but poorly explained: Acknowledge the user's confusion is justified, then re-explain with better evidence. Show the receipts — point to specific code, docs, or behavior that supports the claim.

   If it's genuinely ambiguous: Lay out both sides honestly. Explain the trade-offs without pretending there's a clear winner when there isn't.

Guidelines

• Actually read the code. Don't re-argue from memory — go look at the source.
• The user is probably right to be suspicious. Approach the re-examination with genuine skepticism of the original response.
• Quote specific code when backing up claims. "Trust me" is not a citation.
• If the original suggestion would cause problems the user noticed, give them credit. "Good catch" is a valid response.
• Never double down on a bad take just to save face.