ReferenceError: classifyHandoffIfNeeded is not defined when Task agent completes

Bug Description

When a Task tool agent completes its work, Claude Code throws ReferenceError: classifyHandoffIfNeeded is not defined. The actual agent work completes successfully, but the task is marked as "failed" due to this error in the completion handler.

Environment

• Claude Code version: 2.1.29
• OS: macOS (Darwin 25.2.0)
• Date: 2026-02-02

Steps to Reproduce

1. Use the Task tool to spawn a subagent (e.g., gsd-executor or gsd-verifier agent types)
2. Agent completes all its work successfully (file edits, bash commands, etc.)
3. When the agent finishes and returns results, error occurs

Expected Behavior

Task completes successfully and returns the agent's results to the parent context.

Actual Behavior

Task is marked as "failed" with error message:

Agent "Execute plan 01.1-09" failed: classifyHandoffIfNeeded is not defined

The agent's actual work (file operations, git commands, etc.) completed successfully — only the result handling fails.

Evidence

From agent output logs, the work completed successfully:

• TypeScript typecheck: 7/7 packages passed
• ESLint: 568 warnings, 0 errors
• Git push: succeeded (56ef1de15..9ddc608c6 dev -> dev)

The error occurs in Claude Code's internal task completion processing, not in the agent's execution.

Investigation

• Searched all user configuration files (~/.claude/) — function not defined there
• Searched plugins directory — function not defined there
• Error is a JavaScript ReferenceError, suggesting a missing import or undefined function in Claude Code's internal codebase

Impact

Low severity — agent work completes, only reporting is affected. Users see "failed" status despite successful execution.

Workaround

None needed for functionality. Users can verify work completed by checking output files directly.