### System details
Intel Core Ultra X7 358H, Intel Arc B390 (Panther Lake), Omarchy 4.0.4-1, codex-cli 0.162.0

### What's wrong?

The `omarchy.agents` bar widget usually shows **"Codex limits unavailable"** for Codex, even though the account is signed in and `codex app-server` answers fine. The record at `~/.local/state/omarchy/agents/usage/codex.json` has `"usageStatusText":"Codex limits unavailable","authHelpText":"account/read"`, meaning `rpc_request` raised `TimeoutError("account/read")`. Claude limits display fine.

**Root cause:** in `bin/omarchy-agent-usage-codex`, `rpc_request()` calls `select()` on `proc.stdout` and then `proc.stdout.readline()`. `proc.stdout` is a buffered `TextIOWrapper`, so `readline()` can pull several lines from the pipe into Python's buffer. Right after `initialized`, codex app-server sends a `remoteControl/status/changed` notification and then the `account/read` response, about 0.5s in and often in a single chunk. `readline()` returns the notification, the response sits in the userspace buffer, `select()` on the fd never reports readable again, and the loop spins until the 4s timeout. It's timing-dependent, so it fails intermittently.

**Repro:**
```bash
for i in 1 2 3 4; do omarchy-agent-usage-codex --force | jq -c '[.limits|length, .usageStatusText]'; done
```
On my machine the stock collector failed 3 out of 4 runs (`[0,"Codex limits unavailable"]`). The patched version below succeeded 4 out of 4 with the same environment (`[2,""]`, plan `plus`, session and weekly limits populated).

**Fix:** read the raw fd with `os.read` and split lines ourselves, so buffered data is consumed before blocking on `select()`. Leftover bytes are kept per fd between calls, so a response that arrives with an earlier one isn't lost.

```diff
--- a/bin/omarchy-agent-usage-codex
+++ b/bin/omarchy-agent-usage-codex
@@ -474,24 +474,34 @@
     return stats
 
 
+_rpc_pending = {}
+
+
 def rpc_request(proc, request_id, method, params=None, timeout=8):
   payload = {"id": request_id, "method": method, "params": params or {}}
   proc.stdin.write(json.dumps(payload) + "\n")
   proc.stdin.flush()
+  fd = proc.stdout.fileno()
+  buf = _rpc_pending.setdefault(fd, b"")
   deadline = time.time() + timeout
   while time.time() < deadline:
-    ready, _, _ = select.select([proc.stdout], [], [], 0.25)
+    while b"\n" in buf:
+      line, buf = buf.split(b"\n", 1)
+      _rpc_pending[fd] = buf
+      try:
+        message = json.loads(line)
+      except Exception:
+        continue
+      if message.get("id") == request_id:
+        return message
+    ready, _, _ = select.select([fd], [], [], 0.25)
     if not ready:
       continue
-    line = proc.stdout.readline()
-    if not line:
+    chunk = os.read(fd, 65536)
+    if not chunk:
       break
-    try:
-      message = json.loads(line)
-    except Exception:
-      continue
-    if message.get("id") == request_id:
-      return message
+    buf += chunk
+    _rpc_pending[fd] = buf
   raise TimeoutError(method)
 
 
```