feat: extract in-process MCP action layer, concurrent task dispatch, and spawn/reconcile hardening

- Extract device action execution into artemis/mcp action layer:
  typed ActionResult, capability manifest with startup validation,
  AdbActuator as the single physical implementation, in-memory
  ActionSession (replaces stdio subprocess MCP in validator), and
  McpActionExecutor preserving the legacy executor contract.
  adb_server keeps its external stdio/SSE surface unchanged and now
  reuses the actuator implementation; import-time side effects moved
  into configure_stdio_mode().
- Capability-gated prompt assembly (prompt_assembly.py): tool enums and
  teaching segments render per available backend tools; full-set output
  is byte-identical to the previous static text (locked by tests).
- Admin console task queue: serial worker replaced by a per-device
  concurrent dispatcher (ARTEMIS_CONCURRENCY_MODE=global restores
  serial semantics), per-session cancellation, strict device binding on
  enqueue, unified stdio piping with tee to trace stdout.log, and
  shutdown cleanup of all active runs. Fix _pid_exists exception order
  so PermissionError is treated as alive.
- mcp_server reliability: spawn watchdog kills runners stuck before
  interpreter startup and releases device reservations; task state
  reconcile uses psutil liveness, prefers runner-reported pids, and
  adds a 45s startup grace to stop false failure reports on Windows;
  conversation_id is now optional end-to-end; explicit device_serial is
  validated at submission.
- CLI install now writes ~/.claude/rules/artemis.md for Claude Code and
  migrates/removes the legacy CLAUDE.md injection block; ~/.claude.json
  is never rewritten when it fails to parse.
- Tests: Windows-portable stdio lifecycle and path assertions; new
  suites for action layer, prompt assembly, spawn watchdog, and task
  reconcile (895 passed, 3 skipped).

mcp_server/server.py

@@ -24,24 +24,49 @@ if PROJECT_ROOT not in sys.path:
     sys.path.insert(0, PROJECT_ROOT)
 
 # Debug logging and stderr tee to capture crashes without breaking MCP stdio JSON-RPC
+_LOG_MAX_BYTES = 5 * 1024 * 1024
+
+
+def _rotate_log(path: str, max_bytes: int = _LOG_MAX_BYTES) -> None:
+    """Rotates the log to <path>.old once it exceeds max_bytes, so long-lived
+    server processes cannot grow the debug logs without bound."""
+    try:
+        if os.path.exists(path) and os.path.getsize(path) > max_bytes:
+            old_path = path + ".old"
+            if os.path.exists(old_path):
+                os.remove(old_path)
+            os.replace(path, old_path)
+    except Exception:
+        pass
+
+
 try:
     _mcp_log_dir = os.path.join(PROJECT_ROOT, "scratch")
     os.makedirs(_mcp_log_dir, exist_ok=True)
 
-    with open(os.path.join(_mcp_log_dir, "mcp_launch_debug.log"), "a", encoding="utf-8") as _f:
+    _launch_log = os.path.join(_mcp_log_dir, "mcp_launch_debug.log")
+    _rotate_log(_launch_log)
+    with open(_launch_log, "a", encoding="utf-8") as _f:
         _f.write(
             f"[{datetime.datetime.now()}] Artemis MCP Server launched! sys.argv: {sys.argv}, CWD: {os.getcwd()}\n"
         )
 
     class StderrTee:
-        def __init__(self, original, log_path):
+        def __init__(self, original, log_path, max_bytes=_LOG_MAX_BYTES):
             self.original = original
+            self.log_path = log_path
+            self.max_bytes = max_bytes
+            _rotate_log(log_path, max_bytes)
             self.log_file = open(log_path, "a", encoding="utf-8")
 
         def write(self, data):
             self.original.write(data)
             self.log_file.write(data)
             self.log_file.flush()
+            if self.log_file.tell() > self.max_bytes:
+                self.log_file.close()
+                _rotate_log(self.log_path, self.max_bytes)
+                self.log_file = open(self.log_path, "a", encoding="utf-8")
 
         def flush(self):
             self.original.flush()