[FEATURE REQUEST] Add subagent output control settings or revert to previous behavior
Feature Request: Add subagent output control settings or revert to previous behavior
Problem
In Claude Code 2.0.14, subagent output (especially from Task tool invocations) is displayed in full detail, which causes severe performance degradation and UI freezing when subagents produce large amounts of output.
Example scenario:
- Using
gemini-clisubagent with web search - Generates hundreds of lines of output
- UI becomes unresponsive and conversation view freezes
- Makes Claude Code unusable for certain tasks
Previous Behavior
In earlier versions, subagent output was not fully visible, as mentioned in #2685. While that issue reported wanting MORE visibility, the current implementation has gone too far in the opposite direction.
Requested Solutions
Option 1: Add output control settings (preferred)
Add configuration options in settings.json:
{
"subagent": {
"outputMode": "summary", // "full" | "summary" | "minimal"
"maxOutputLines": 50
}
}
Option 2: Revert to previous behavior
Return to the previous implementation where subagent execution happened silently, showing only final results.
Why This Matters
- Performance: Large output causes UI freezing and conversation lag
- Usability: Users cannot effectively work with verbose subagents
- Stability: Current behavior makes certain legitimate use cases impossible
Additional Context
- Version: 2.0.14
- Related issues: #2685 (opposite problem), #3217 (verbose setting removal)
- The
verbosesetting was removed in recent versions, leaving no way to control output verbosity
Proposed Workaround
Until this is fixed, the only option is to avoid using subagents that produce large output, which severely limits Claude Code's usefulness.
This issue has 4 comments on GitHub. Read the full discussion on GitHub ↗