diff --git a/bin/gstack-codex-probe b/bin/gstack-codex-probe
index 772d2487e..dcc9e3bc2 100755
--- a/bin/gstack-codex-probe
+++ b/bin/gstack-codex-probe
@@ -418,6 +418,9 @@ _gstack_codex_timeout_wrapper() {
         sleep 1
         _GSTACK_CODEX_WAITED=$((_GSTACK_CODEX_WAITED + 1))
       done
+      # Freeze the command first: killing its children while it runs lets it
+      # continue past the killed child and print output after its deadline.
+      kill -STOP "$_cmd_pid" 2>/dev/null
       pkill -KILL -P "$_cmd_pid" 2>/dev/null
       kill -KILL "$_cmd_pid" 2>/dev/null
       exit 124
diff --git a/test/codex-hardening.test.ts b/test/codex-hardening.test.ts
index bb345a4a5..c0e8efdc4 100644
--- a/test/codex-hardening.test.ts
+++ b/test/codex-hardening.test.ts
@@ -357,7 +357,10 @@ describe('gstack-codex-probe: timeout wrapper + namespace hygiene', () => {
     // tool gate killed the whole call and the partial output was lost.
     const dir = fs.mkdtempSync(path.join(os.tmpdir(), 'gstack-watchdog-stubborn-'));
     try {
-      for (const tool of native ? ['bash', 'sleep', 'cat'] : []) {
+      // pkill is part of the watchdog's contract (it reaps the provider's
+      // children; stock macOS ships /usr/bin/pkill). Without it the orphaned
+      // `sleep 30` keeps stdout open and the capture waits out the sleep.
+      for (const tool of native ? ['bash', 'sleep', 'cat', 'pkill'] : []) {
         const resolved = spawnSync('bash', ['-c', `command -v ${tool}`], { timeout: 5000 });
         fs.symlinkSync(resolved.stdout.toString().trim(), path.join(dir, tool));
       }