[Bug] Anthropic API Error: False positive safeguard flags on claude-opus-5-5 with reasoning_extraction

Bug Description
False positive safeguard flags on Opus 5.5 (claude-opus-5-5), detail [reasoning_extraction]. Context: normal software project (children's events map). The orchestrator asked background subagents to quote the "You are powered by the model named..." line from their context, only to verify they run on Opus 5.5. Every subagent then failed with "safeguards flagged this message", and each resume failed again because the request stayed in the transcript. Nothing harmful was requested; the work was a read-only code and server audit of our own project. We have stopped doing this. Please confirm our account is not at risk. Request IDs: req_011CfK5w99Yj5V3HW9KSrjHR, req_011CfK5wy37mjqCa5d8ATUJM, req_011CfK5yTk3Zxx6kfmzHgLtK, req_011CfK5zHEoeeHwi4w6JJJrb

Environment Info

• Platform: win32
• Terminal: vscode
• Version: 2.1.280
• Feedback ID: d0a69339-54bb-4732-9fa4-6902471eeed3

Errors

[]