[BUG] Repeated failures and token waste when parsing dpu.kev under QNX 8.0 using Claude terminal

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Using Claude's terminal, the first input was: Provide a dpu.kev file under QNX 8.0, specify the QNX 8.0 toolchain, and parse it with traceprinter. It didn't find the answer and just gave a random explanation.
Then I recaptured the dpu.kev file for 60 seconds and asked Claude to parse it again. This time it completely ignored my command and still performed the same old analysis.
There are many similar issues. This is unbearable and wastes a lot of tokens

<img width="2171" height="1071" alt="Image" src="https://github.com/user-attachments/assets/03e0c483-5487-4b8e-8b37-b479a58c1cee" />

What Should Happen?

Expected behavior: Any analysis must be rigorous and well-founded, not based on random guesses. At the same time, it must strictly follow user instructions, rather than running wild according to its own logic

Error Messages/Logs

Steps to Reproduce

Provide a dpu.kev captured on QNX 8.0 using tracelogger (approximately 20 MB), then ask Claude to analyze the specific process creation events.

Claude analyzes it, but goes off track.

Following Claude’s suggestion, capture the trace again, re-upload it, and ask Claude to perform a new analysis. However, in the same conversation window, Claude does not perform a new analysis at all — it just repeats the old one.

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.160 (Claude Code)

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

Other

Additional Information

_No response_