fix(shell): interruptible timeout + kill the whole process tree (#62)

withTimeout wrapped a blocking process.waitFor() that coroutine cancellation
can't interrupt, so a hung process hung the stage until it exited on its own;
and destroyForcibly() killed only the direct child, leaving sh -c grandchildren
(the real npm/tsc) as orphans. Use process.waitFor(timeout, MILLISECONDS) for
an on-the-clock deadline, and kill via toHandle().descendants() + the parent.

infrastructure/tools/src/main/kotlin/com/correx/infrastructure/tools/shell/ShellTool.kt

@@ -16,10 +16,9 @@ import com.correx.core.tools.contract.ToolResult
 import com.correx.core.tools.contract.ValidationResult
 import com.correx.core.tools.process.ChildProcess
 import kotlinx.coroutines.Dispatchers
-import kotlinx.coroutines.TimeoutCancellationException
 import kotlinx.coroutines.async
+import kotlinx.coroutines.coroutineScope
 import kotlinx.coroutines.withContext
-import kotlinx.coroutines.withTimeout
 import kotlinx.serialization.json.Json
 import kotlinx.serialization.json.JsonObject
 import kotlinx.serialization.json.JsonPrimitive
@@ -28,6 +27,7 @@ import kotlinx.serialization.json.buildJsonObject
 import kotlinx.serialization.json.put
 import kotlinx.serialization.json.putJsonObject
 import java.nio.file.Path
+import java.util.concurrent.TimeUnit
 
 class ShellTool(
     private val allowedExecutables: Set<String> = emptySet(),
@@ -195,21 +195,13 @@ class ShellTool(
                 runCatching {
                     runCmd(request, process)
                 }.getOrElse {
-                    process.destroyForcibly()
+                    killTree(process)
-                    if (it is TimeoutCancellationException) {
-                        ToolResult.Failure(
-                            invocationId = request.invocationId,
-                            reason = "Process timed out after ${timeoutMs}ms",
-                            recoverable = false,
-                        )
-                    } else {
                     ToolResult.Failure(
                         invocationId = request.invocationId,
                         reason = it.message ?: "Unknown error occurred during execution",
                         recoverable = false,
                     )
                 }
-                }
             } ?: ToolResult.Failure(
                 invocationId = request.invocationId,
                 reason = (this as ValidationResult.Invalid).reason,
@@ -254,11 +246,32 @@ class ShellTool(
             .joinToString("\n")
     }
 
-    private suspend fun runCmd(request: ToolRequest, process: Process): ToolResult = withTimeout(timeoutMs) {
+    // Kill the whole process tree. destroyForcibly() signals only the direct child, so an `sh -c`
-        val stdoutDeferred = async { process.inputStream.bufferedReader().use { it.readText() } }
+    // command's grandchildren (the actual npm/tsc/…) would survive as orphans; descendants() reaches
-        val stderrDeferred = async { process.errorStream.bufferedReader().use { it.readText() } }
+    // them. Snapshot descendants BEFORE destroying the parent — reparenting can hide them afterwards.
+    private fun killTree(process: Process) {
+        process.toHandle().descendants().forEach { it.destroyForcibly() }
+        process.destroyForcibly()
+    }
 
-        val exitCode = process.waitFor()
+    private suspend fun runCmd(request: ToolRequest, process: Process): ToolResult = coroutineScope {
+        val stdoutDeferred = async(Dispatchers.IO) { process.inputStream.bufferedReader().use { it.readText() } }
+        val stderrDeferred = async(Dispatchers.IO) { process.errorStream.bufferedReader().use { it.readText() } }
+
+        // waitFor(timeout) enforces the deadline on the clock — the plain blocking waitFor() is not
+        // coroutine-cancellable, so a withTimeout around it can't actually interrupt a hung process.
+        val finished = withContext(Dispatchers.IO) { process.waitFor(timeoutMs, TimeUnit.MILLISECONDS) }
+        if (!finished) {
+            killTree(process)
+            stdoutDeferred.cancel()
+            stderrDeferred.cancel()
+            return@coroutineScope ToolResult.Failure(
+                invocationId = request.invocationId,
+                reason = "Process timed out after ${timeoutMs}ms",
+                recoverable = false,
+            )
+        }
+        val exitCode = process.exitValue()
         val stdout = stdoutDeferred.await()
         val stderr = stderrDeferred.await()