fix: orchestrator exits immediately when using Claude Code harness #3
No reviewers
Labels
No labels
bug
documentation
duplicate
enhancement
good first issue
help wanted
invalid
question
wontfix
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
backmeupplz/superharness!3
Loading…
Reference in a new issue
No description provided.
Delete branch "fix/claude-orchestrator-exit"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Problem
When Claude Code is the selected harness, superharness exits immediately after the orchestrator processes its initial prompt. The user sees:
Root Cause
Both the orchestrator (
session.rs) and workers (panes.rs) use the samebuild_harness_cmdfunction, which passes--printto Claude Code. The--printflag runs Claude in non-interactive mode — it processes the prompt, prints output, and exits.This is correct for workers (run a task and exit), but wrong for the orchestrator which needs to stay alive interactively for the human to communicate with it.
Fix
Added a new
build_orchestrator_cmdfunction inharness.rsthat omits--printfor Claude Code, passing the prompt as a positional argument instead. This keeps the orchestrator session interactive.build_harness_cmdclaude --dangerously-skip-permissions --print <prompt>build_orchestrator_cmdclaude --dangerously-skip-permissions <prompt>Workers continue to use
build_harness_cmdwith--printas intended.Changes
src/harness.rs: Addedbuild_orchestrator_cmdfor interactive orchestrator sessionssrc/tmux/session.rs: Orchestrator init now callsbuild_orchestrator_cmdinstead ofbuild_harness_cmdThanks for the PR! This was already fixed in main — we split the orchestrator and worker command paths so the orchestrator no longer gets
--printwhen using Claude Code. Closing as already resolved, but really appreciate the clean write-up and diagnosis!Pull request closed