fix: address codex+gpt-5.5 review findings

- harvest: tighten sub-3s filter to also require prompt < 200 chars,
  avoiding false positives on fast real one-shot questions
- openclaw schedule_cmd: add docstring clarifying it schedules the
  shared engine, not the OpenClaw-native runner

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

plugins/openclaw/slash_sleep.py

@@ -208,7 +208,12 @@ def reject(night: str = None) -> int:
 
 
 def schedule_cmd(hour: int, minute: int) -> int:
-    """Install a nightly cron entry via the shared SkillOpt-Sleep scheduler."""
+    """Install a nightly cron entry via the shared SkillOpt-Sleep scheduler.
+
+    Note: this schedules the shared engine (``python -m skillopt_sleep run``),
+    not the OpenClaw-specific ``run_sleep.py``.  Use ``run_sleep_cron.sh`` if
+    you need the OpenClaw-native backend and category task files instead.
+    """
     try:
         from skillopt_sleep.scheduler import schedule
     except ImportError:

skillopt_sleep/harvest.py

@@ -148,8 +148,11 @@ def _is_headless_replay(digest: "SessionDigest") -> bool:
     for marker in _REPLAY_PROMPT_MARKERS:
         if marker in prompt:
             return True
-    # Sub-3-second single-turn sessions are almost certainly programmatic.
+    # Sub-3-second single-turn sessions with short prompts are almost
-    if digest.started_at and digest.ended_at:
+    # certainly programmatic (engine grader/judge calls).  We require the
+    # prompt to also be short (<200 chars) to avoid false-positives on
+    # real one-shot questions that Claude happens to answer quickly.
+    if digest.started_at and digest.ended_at and len(prompt) < 200:
         try:
             fmt = "%Y-%m-%dT%H:%M:%S"
             start = datetime.strptime(digest.started_at[:19], fmt)