Environment

• OS: Windows 11 Pro 10.0.26200
• Claude Code: 2.1.109
• Shell: Git Bash
• Node: v24.14.0
• codex plugin: openai-codex/codex v1.0.2 (installed 2026-04-04)
• codex CLI: 0.118.0 (npm global)

Summary

Invoking /codex:rescue, /codex:review, /codex:adversarial-review through the skill dispatch hangs indefinitely about 20% of the time. The turn never completes — no output, no error, no timeout. Direct invocation of codex exec from Git Bash always works.

Observed Behavior

• Skill dispatch: ~80% succeed normally, ~20% hang for >20 minutes with no output
• No stderr, no recovery — user must abandon the turn
• Happens across all three commands (rescue, review, adversarial-review)
• Over one day of normal use: 80+ invocations, 18+ hangs

Forensic Evidence

After a hang, the following zombie state is left behind:

1. Stale broker directory /tmp/cxc-<id>/ persists with:
   • broker.pid containing a PID that is no longer alive
   • broker.log containing only deprecation warnings, no IPC activity
   • No *.sock file (socket never bound, or cleaned on crash)
2. Empty state directory /tmp/codex-companion/<workspace-hash>/ exists but jobs/ was never created — indicating writeJobFile() was never reached.
3. PID in broker.pid is a genuine zombie — kill -0 $PID returns failure, but directory is not cleaned.

Root Cause Analysis

Tracing through the plugin source (paths under ~/.claude/plugins/cache/openai-codex/codex/1.0.2/):

User /codex:rescue
  → skill wrapper (rescue.md)
    → node codex-companion.mjs task
      → withAppServer(cwd, fn)                    [codex.mjs:607]
        → CodexAppServerClient.connect(cwd)        [codex.mjs:610]
          → ensureBrokerSession(cwd)               [broker-lifecycle.mjs:113]
            → spawnBrokerProcess()                 [broker-lifecycle.mjs:59-70]
            → waitForBrokerEndpoint(2000ms)        [broker-lifecycle.mjs:149]
              ⨯ HANGS HERE on Windows

Hypotheses (Windows-specific)

1. Unix socket path emulation: brokerEndpoint at /tmp/cxc-*/*.sock is created under Windows' emulated /tmp. Under Node.js on Windows, unix socket connection is unreliable — especially when the broker child process is spawned with detached: true, stdio: ["ignore", logFd, logFd] (broker-lifecycle.mjs:59-70). File descriptor writes are not atomic on Windows.
2. Line buffering + CRLF: handleLine() in app-server.mjs:117-128 splits JSON-RPC messages on \n. If the broker emits \r\n on Windows, upstream parsers may drop notifications — including the completion signal.
3. Race condition in scheduleInferredCompletion() (codex.mjs:367-388): uses a 250ms setTimeout to detect "inferred completion" against state.pendingCollaborations and state.activeSubagentTurns. A missed notification + stale set membership → turn never completes → captureTurn() (codex.mjs:553-605) never resolves.
4. Broker race on cold start: first connection after a new broker spawn is where the 20% failure clusters. Warm broker (previous task succeeded) tends to stay stable.

Minimal Repro Attempt

No reliable repro — it's intermittent. Rough steps:

1. Leave codex plugin idle for 30+ minutes (broker dies)
2. Invoke /codex:rescue "any prompt"
3. ~1 in 5 invocations hangs indefinitely; rest complete normally

Workaround

Bypass the skill entirely and call codex exec directly from bash. Example wrapper:

```bash

direct codex exec + zombie cleanup

shopt -s nullglob for dir in /tmp/cxc-*; do PID=$(cat "$dir/broker.pid" 2>/dev/null) if [[ -n "$PID" ]] && kill -0 "$PID" 2>/dev/null; then continue; fi AGE_MIN=$(( ( $(date +%s) - $(stat -c %Y "$dir") ) / 60 )) [[ $AGE_MIN -ge 30 ]] && rm -rf "$dir" done codex exec "$PROMPT" ```

This is stable across 100+ invocations, confirming the issue is in the broker/IPC layer, not codex CLI itself.

Requested Fix Direction

1. Treat waitForBrokerEndpoint() timeout as fatal — surface an error to Claude Code instead of hanging Promise forever
2. Add CRLF normalization in handleLine() before JSON-RPC parse
3. On Windows, fall back from unix socket to named pipe (\.\pipe\codex-*) which is native Windows IPC and better supported by Node.js
4. Add pre-flight cleanup of stale /tmp/cxc-* directories in ensureBrokerSession()

Impact

For users running Claude Code on Windows, this silently burns 20% of codex-related interactions. Hard to diagnose without plugin internals, and no user-visible error means users often blame their own prompts.