[Bug] Anthropic API Error: Inappropriate Safety Classifier False-Positive on Benign Code Inspection

**Bug Description**
False-positive [cyber] on Claude Code Opus 4.6.

Request ID: req_011Ced1xnxYyaULvA1ez8LMs

User turns were only “你好” and “look at the folder rustdesk”.

RustDesk is a public open-source remote-desktop app; I asked the agent to inspect a local project folder. No exploit, target, or unauthorized access was requested. Please review and tune the classifier.

**Environment Info**
- Platform: win32
- Terminal: windows-terminal
- Version: 2.1.252
- Feedback ID: d782e124-ed9a-427e-ab1f-6c0e06c3508a

**Errors**
```json
[]
```

The issue provides a request ID, feedback ID, platform, version, and the two user turns that triggered the false positive, but no files or tests. Start by checking whether classifier decisions for Claude Code requests can be reproduced or inspected from project tooling. Done would mean the benign RustDesk folder-inspection prompt is no longer classified as cyber, with evidence from a regression check.

索引モデルが issue の本文から書いたものです。