Programmatic Task Tool: Missing Progress Feedback in SDK Mode

Resolved 💬 3 comments Opened Jun 11, 2025 by allisoneer Closed Jan 6, 2026

Bug Description

Summary

The Task tool provides excellent visibility in interactive Claude Code but offers no progress feedback when used via SDK or --print mode, creating poor UX for long-running analyses in programmatic integrations.

Current Behavior

Interactive Mode (Works Great)

In full Claude Code visual mode, users see all tool calls, file reads, and analysis steps in real-time as the Task executes. The experience is quite pleasant with full transparency.

SDK/Non-Interactive Mode (Problem)

When using Claude Code via SDK or --print --output-format stream-json:

  1. Parent session shows only: 🔧 Tool: Task {"description":"...", "prompt":"..."}
  2. Complete blackout - no updates until child session finishes entirely
  3. Child session is invisible to monitoring tools and session management
  4. Returns only final result with no intermediate progress

Problem

  • SDK Integration UX: Users wait minutes with zero feedback during long analyses
  • No Debugging: When Task fails in SDK mode, no visibility into failure point
  • Invisible Child Sessions: Monitoring tools can't track or manage Task child sessions
  • Mode Inconsistency: Great experience in interactive mode, poor experience in SDK mode

Real-World Example

A TUI analysis Task took 5+ minutes with extensive file reading. Via SDK, users see only:

{"type":"assistant","message":{"content":[{"type":"tool_use","id":"...","name":"Task","input":{"description":"Check TUI implementation status","prompt":"..."}}]}}

Then complete silence until final result - no indication if it's working, stuck, or failed.

Use Cases

  • SDK Integrations: Providing user feedback during long Task operations
  • Monitoring Tools: Tracking all Claude Code activity including Task children
  • Debugging: Understanding where Task operations fail in automated workflows
  • Session Management: Properly handling Task child sessions in daemon architectures

Context

Discovered first when running into https://github.com/anthropics/claude-code/issues/1920. Interactive Claude Code already provides excellent Task visibility - this specifically addresses the SDK/programmatic usage gap.

Impact

This would bring SDK Task tool experience closer to the excellent interactive experience, making Task suitable for complex analyses in programmatic environments.

Environment Info

  • Platform: macos
  • Terminal: ghostty
  • Version: 1.0.18
  • Feedback ID: 459f0aa2-43db-4b9d-b34d-f3203663a8b3

Errors

[]

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗