[BUG] Dispatch: reasoning_extraction safeguard fires on bare 'hello' since Sept 1

### Preflight Checklist

- [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) 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?

Dispatch (Desktop, macOS) has returned "Opus 5's safeguards flagged this message
[reasoning_extraction]" on a bare "hello" since Sept 1. Request IDs:
req_011CeeiaWbU2833wn9TUejN5 (Sept 2), req_011CerCfF6kTMpxHxEbLsZNi (Sept 8 11:22),
req_011CerDgPZPQ1psNiAzL99Ys (Sept 8 11:35). Steps taken with no effect: project
CLAUDE.md reduced from 1.6MB to 28KB; every instruction mentioning "reasoning" or
"memory" reworded; app quit and reopened; Cowork storage moved to recommended
location; Notion reconnected; repo pulled clean; new conversations started. Regular
Claude Code cloud sessions on the same repo work. Dispatch has no model selector.
Separately, since Sept 6 both Cowork and Claude Code have started sessions with an
empty "continued from a previous conversation" summary. Account: Max. This blocks a
production workflow.

### What Should Happen?

Dispatch (Desktop, macOS) has returned "Opus 5's safeguards flagged this message
[reasoning_extraction]" on a bare "hello" since Sept 1. Request IDs:
req_011CeeiaWbU2833wn9TUejN5 (Sept 2), req_011CerCfF6kTMpxHxEbLsZNi (Sept 8 11:22),
req_011CerDgPZPQ1psNiAzL99Ys (Sept 8 11:35). Steps taken with no effect: project
CLAUDE.md reduced from 1.6MB to 28KB; every instruction mentioning "reasoning" or
"memory" reworded; app quit and reopened; Cowork storage moved to recommended
location; Notion reconnected; repo pulled clean; new conversations started. Regular
Claude Code cloud sessions on the same repo work. Dispatch has no model selector.
Separately, since Sept 6 both Cowork and Claude Code have started sessions with an
empty "continued from a previous conversation" summary. Account: Max. This blocks a
production workflow.

### Error Messages/Logs

```shell
API Error: Opus 5's safeguards flagged this message (https://www.anthropic.com/legal/aup). This sometimes happens with safe, normal conversations. Claude Code can't respond to this message with Opus 5.

Try rephrasing the request in a new session or change your model.

Learn more: https://support.claude.com/en/articles/16049681

Details: [reasoning_extraction]

Request ID: req_011CerDgPZPQ1psNiAzL99Ys
```

### Steps to Reproduce

just say hello to dispatch

### Claude Model

Sonnet (default)

### Is this a regression?

Yes, this worked in a previous version

### Last Working Version

_No response_

### Claude Code Version

2.1.12 (Claude Code)

### Platform

Anthropic API

### Operating System

macOS

### Terminal/Shell

Terminal.app (macOS)

### Additional Information

_No response_

No repository file, test, or entry point is named. Start by reproducing the bare "hello" in Dispatch on macOS and reviewing the reported reasoning_extraction error and request IDs. Done means the safeguard no longer blocks a harmless message and the empty continuation summary is accounted for, with regression coverage if the repository contains the relevant path.

Written by the indexing model from the issue text.