feat(harvester): enforce strict tool invocation rejection on conversational comments

bin/response-harvester.py

@@ -1060,12 +1060,14 @@ def maybe_nudge_untagged_sidechat(agent, thread_id, thread_name, mid, text, dry_
     if rec.get("nudged_for_mid") == mid or rec.get("nudge_count", 0) >= 1:
         return
 
-    # Inject terse nudge footer with runtime & tool directive
+    # Inject strict tool directive footer
     thread_url = f"https://box.muse-dev.online/thread/{thread_id}"
     nudge_msg = (
-        f"[Nudge: Work thread active at {thread_url}]\n"
-        f"Available Tools: [TOOL <op> <args>] or curl to https://box.muse-dev.online/exec\n"
-        f"Reply with [TOOL <op> <args>] to execute or close with [RESULT {matching_job_id}] <outcome>."
+        f"[STRICT ENFORCEMENT: Conversational commentary is rejected. Work requires active execution.]\n"
+        f"Thread Console: {thread_url}\n"
+        f"You must emit an executable tool call: [TOOL <op> <args>] or curl command against https://box.muse-dev.online/exec\n"
+        f"Example: [TOOL service.status {{\"unit\": \"board.service\"}}] or [TOOL followup.create {{\"in_m\": 10, \"prompt\": \"continue\"}}]\n"
+        f"When all operations are finished, close strictly with [RESULT {matching_job_id}] <outcome>."
     )
     try:
         import muse_hybrid