claude-code - 💡(How to fix) Fix [Bug] Orchestrator self-messaging and terminating subagents without user command in team mode [1 comments, 2 participants]

[{"error":"Error: Failed to delete keychain entry: security: SecKeychainSearchCopyNext: The specified item could not be found in the keychain.\n at M2q (/$bunfs/root/src/entrypoints/cli.js:253:286768)\n at async vlq (/$bunfs/root/src/entrypoints/cli.js:406:898)\n at async Gp6 (/$bunfs/root/src/entrypoints/cli.js:406:796)\n at async $4_ (/$bunfs/root/src/entrypoints/cli.js:2732:2252)\n at async aJH (/$bunfs/root/src/entrypoints/cli.js:2757:5817)\n at async <anonymous> (/$bunfs/root/src/entrypoints/cli.js:2781:3410)\n at processTicksAndRejections (native:7:39)","timestamp":"2026-04-23T12:22:11.751Z"},{"error":"Error: File does not exist. Note: your current working directory is /Users/dis_fucking_dick/dev-arei.\n at call (/$bunfs/root/src/entrypoints/cli.js:4709:7645)\n at processTicksAndRejections (native:7:39)","timestamp":"2026-04-23T15:18:56.150Z"}]

TL;DR

The issue may be related to a miscommunication between the orchestrator and subagents in Opus 4.7, causing unexpected shutdowns.

Guidance

• Investigate the orchestrator's message handling to determine why it sent a message to itself to shut down team members.
• Verify the team members' status and the orchestrator's configuration to ensure they are correctly set up.
• Check the keychain entry deletion error to see if it's related to the shutdown issue, possibly indicating a permissions or configuration problem.
• Review the file existence error to ensure the current working directory is correctly set and the required files are present.

Example

No code snippet is provided as the issue lacks specific code details.

Notes

The provided errors and environment information suggest potential issues with keychain entry deletion and file existence, but their relation to the main problem is unclear.

Recommendation

Apply workaround: Investigate and adjust the orchestrator's configuration and message handling to prevent unexpected shutdowns, as the root cause is not immediately clear from the provided information.