diff --git a/user/sigchldtest.c b/user/sigchldtest.c
new file mode 100644
index 0000000..313cbda
--- /dev/null
+++ b/user/sigchldtest.c
@@ -0,0 +1,80 @@
+#include "ulib.h"
+
+/*
+ * Boot-time driver for real, kernel-generated SIGCHLD (sched.h's new
+ * Thread.parentId, threadExit/threadKill's new sigRaise(parentId, sigChld)
+ * call): proves the kernel itself notifies a process's SPAWNING parent the
+ * instant a child it created via sysSpawn actually dies, asynchronously,
+ * without that parent ever having to call sysWait first to find out.
+ *
+ * eintrtest.c already proves early-wake delivery works generically for an
+ * arbitrary signal explicitly sent at a controlled time; proctest.c
+ * already proves sysWait/threadJoin's synchronous "block until this pid is
+ * Dead" contract. Neither proves the ONE thing new here: that the kernel
+ * itself, unprompted, raises a specific signal (sigChld) against a
+ * specific target (the spawning parent, not just "whoever calls wait")
+ * from exactly the right lifecycle event (threadExit/threadKill).
+ *
+ * Uses the same bounded "check a flag every userSleep(20)" poll idiom
+ * masktarget.c/nesttarget.c already use for their own outer verification,
+ * deliberately NOT eintrtarget.c's tight write-then-block-forever shape:
+ * sigchldchild.c's death is not explicitly time-controlled by any
+ * userSigSend call the way eintrtest.c's sigTerm or masktest.c's sigUsr1/
+ * sigUsr2 are, so there is no readiness rendezvous to build here that
+ * would not itself just be racing the same autonomous kernel event it is
+ * trying to order against. The poll loop sidesteps that entirely: whether
+ * sigchldchild.c dies before or after any given iteration starts, the very
+ * next natural sleep-wake cycle's return to ring 3 re-checks pending
+ * signals (isr.c's deliverPendingSignals, unconditional on every return),
+ * so the pending sigChld bit is picked up regardless of ordering -- no
+ * timing guess, just a different (still real, still ring-3) rendezvous
+ * shape than the tight-block one.
+ */
+
+static volatile int handlerFired = 0;
+
+static void onChld(int signum)
+{
+    (void)signum;
+    print("sigchldtest: sigchld-handler-fired\n");
+    handlerFired = 1;
+}
+
+int main(void)
+{
+    print("sigchldtest: start, pid=");
+    printDec(userGetpid());
+    print("\n");
+
+    if (userSigAction(sigChld, onChld, 0, 0) == (long)sigSetFailed) {
+        print("sigchldtest: FAIL install sigchld handler\n");
+        return 1;
+    }
+
+    long child = userSpawn("/bin/sigchldchild", 0);
+    if (child < 0) {
+        print("sigchldtest: FAIL spawn sigchldchild\n");
+        return 1;
+    }
+
+    for (int i = 0; i < 80 && !handlerFired; i++)
+        userSleep(20);
+
+    if (!handlerFired) {
+        print("sigchldtest: FAIL sigchld handler never fired\n");
+        return 1;
+    }
+    print("sigchldtest: sigchld-delivered-to-parent-ok\n");
+
+    long status = userWait(child);
+    print("sigchldtest: sigchldchild exit status=");
+    printDec(status);
+    print("\n");
+    if (status != 84) {
+        print("sigchldtest: FAIL expected sigchldchild exit status 84\n");
+        return 1;
+    }
+    print("sigchldtest: chld-reap-after-signal-ok\n");
+
+    return 0;
+}