fix: register worker task name at startup to prevent spurious early exit (#165)

The watchWorkerInstancesLoop captures `me` with task=undefined at startup
because the initial getWorkerInstance() call has no task parameter. Any other
process that also calls getWorkerInstance() without a task (e.g. updateGithubActionTask
scheduler running at the same time) matches the task===undefined condition and
incorrectly evicts the main bot process.

Fix: call getWorkerInstance("ComfyPR Bot Running") before Promise.all so that
me.task is already set when watchWorkerInstancesLoop starts. Only a competing
instance with the same task name can now trigger an exit.

Co-authored-by: snomiao <snomiao@gmail.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>

Author: 3 people

src/index.ts

@@ -7,11 +7,16 @@ import { updateAuthors } from "./updateAuthors";
 import { updateCNRepos } from "./updateCNRepos";
 import { runFollowRuleSet } from "./updateFollowRuleSet";
 import { updateSlackMessages } from "./updateSlackMessages";
+import { getWorkerInstance } from "./WorkerInstances";
 import { tLog } from "./utils/tLog";
 
 const logger = createLogger("main");
 
 if (import.meta.main) {
+  // Register task name immediately so watchWorkerInstancesLoop captures me.task correctly.
+  // Without this, me.task is undefined and any other process calling getWorkerInstance()
+  // without a task name would incorrectly evict this process.
+  await getWorkerInstance("ComfyPR Bot Running");
   await Promise.all([
     // try send msgs that didn't send in last run
     updateSlackMessages(),