[Bug] Anthropic API Error: Opus 5 reasoning_extraction block persists across sessions

Bug Description
Bug: Opus 5 blocked with [reasoning_extraction] since Friday, persists across new sessions What happened: I was working on a project using Claude Code with 3 subagents. I asked them to explore new approaches and exchange their reasoning/thinking with each other directly, without printing it as normal output tokens. On my next message, Opus 5 blocked the request with: "API Error: Opus 5's safeguards flagged this message. Details: [reasoning_extraction]" The problem: - This is only affecting Opus (Sonnet and other models work fine). - The block has persisted since Friday until today (several days), even though I've closed the session and started new ones. - This does not match the usual false-positive pattern reported elsewhere, where switching models or removing the flagged content in the same session resolves it immediately. What I need: Could you check whether this is an account-level restriction rather than a per-message flag? I'd like to know if this needs manual review, and how to get Opus access restored. Thanks.

Environment Info

• Platform: win32
• Terminal: windows-terminal
• Version: 2.1.270
• Feedback ID: 30400ab2-bd20-49b3-98a6-c10d884794fb

Errors

[]