[MODEL] Model dismisses valid contextual questions during active development session

Preflight Checklist

• [x] I have searched existing issues for similar behavior reports
• [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Other unexpected behavior

What You Asked Claude to Do

After completing a feature implementation (adding template-based message handling to re-engage conversations disabled by WhatsApp Business API inactivity timeout), I asked a directly related architectural question about how WhatsApp Business API account association works across devices.

What Claude Actually Did

Claude Code responded:

"This is a question about how WhatsApp Business API works, not about the code."

Expected Behavior

Claude Code should either:

• Answer the question (it's relevant to the development context), or
• Say it's not certain and suggest where to find the answer

Files Affected

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

Yes, every time with the same prompt

Steps to Reproduce

1. Work on a feature involving a third-party API (in this case WhatsApp Business API)
2. Complete the implementation
3. Ask a contextual/architectural question about how that API behaves at runtime
4. Observe that the model dismisses the question instead of answering or admitting uncertainty

Claude Model

Sonnet

Relevant Conversation

Impact

Critical - Data loss or corrupted project

Claude Code Version

2.1.123

Platform

Anthropic API

Additional Context

This response is unhelpful and condescending. It does not answer the question, does not say it doesn't know, and implies the user made a mistake by asking it — when in fact the question is entirely relevant to the feature just built.

What it should NOT do is tell the user what their question is or isn't about.

This behavior feels patronizing and breaks the flow of an active development session. Developers regularly ask questions that mix code, architecture, and third-party API behavior — that's normal software development. A tool that polices question categories instead of helping is counterproductive.