diff --git a/strix/core/execution.py b/strix/core/execution.py
index 4bcd31d885..2df8ffb5c0 100644
--- a/strix/core/execution.py
+++ b/strix/core/execution.py
@@ -24,6 +24,7 @@
 
 from strix.config import codex
 from strix.core.hooks import (
+    WRAP_UP_AFTER_TURNS_KEY,
     BudgetExceededError,
     BudgetPausedError,
     SubagentBudgetReservedError,
@@ -39,6 +40,7 @@
 )
 from strix.llm import request_log
 from strix.llm.compaction import is_context_overflow, maybe_compact
+from strix.skills import skill_wrap_up_after_turns
 
 
 if TYPE_CHECKING:
@@ -434,6 +436,7 @@ async def spawn_child_agent(
             task=task,
             parent_history=parent_history,
         ),
+        wrap_up_after_turns=skill_wrap_up_after_turns(skills),
         event_sink=event_sink,
         hooks=hooks,
     )
@@ -512,6 +515,7 @@ async def respawn_subagents(
                 task=str(md.get("task", "")),
                 initial_input=[],
                 start_parked=start_parked,
+                wrap_up_after_turns=skill_wrap_up_after_turns(child_skills),
                 event_sink=event_sink,
                 hooks=hooks,
             )
@@ -1163,6 +1167,7 @@ async def _start_child_runner(
     task: str,
     initial_input: Any,
     start_parked: bool = False,
+    wrap_up_after_turns: int | None = None,
     event_sink: StreamEventSink | None = None,
     hooks: RunHooks[dict[str, Any]] | None = None,
 ) -> None:
@@ -1174,6 +1179,7 @@ async def _start_child_runner(
     child_ctx["agent_id"] = child_id
     child_ctx["parent_id"] = parent_id
     child_ctx["task"] = task
+    child_ctx[WRAP_UP_AFTER_TURNS_KEY] = wrap_up_after_turns
 
     async def _child_loop() -> None:
         # A budget stop is a clean scan-wide shutdown, not a child failure: the
diff --git a/strix/core/hooks.py b/strix/core/hooks.py
index 13fc98f988..6b50ba1c45 100644
--- a/strix/core/hooks.py
+++ b/strix/core/hooks.py
@@ -22,6 +22,9 @@
 
 
 LLM_TURN_KEY = "llm_turn"
+# Set on a sub-agent's context from its skills (see ``skill_wrap_up_after_turns``).
+WRAP_UP_AFTER_TURNS_KEY = "wrap_up_after_turns"
+_WRAP_UP_REPEAT_TURNS = 5
 
 # ``BudgetPolicy`` decides what happens when the accumulated LLM cost reaches
 # ``max_budget_usd``.
@@ -35,6 +38,7 @@
 # same call when the operator raises the limit or resumes.
 __all__ = [
     "LLM_TURN_KEY",
+    "WRAP_UP_AFTER_TURNS_KEY",
     "BudgetExceededError",
     "BudgetPausedError",
     "BudgetPolicy",
@@ -207,6 +211,7 @@ async def on_llm_start(
         context.context[LLM_TURN_KEY] = int(context.context.get(LLM_TURN_KEY, 0)) + 1
         try:
             self._maybe_warn_turns(context, input_items)
+            self._maybe_nudge_wrap_up(context, input_items)
             self._maybe_warn_budget(context, input_items)
         except Exception:
             logger.exception("budget/turn warning injection failed")
@@ -270,6 +275,26 @@ def _maybe_warn_turns(
         )
         input_items.append({"role": "user", "content": content})
 
+    def _maybe_nudge_wrap_up(
+        self,
+        context: RunContextWrapper[dict[str, Any]],
+        input_items: list[TResponseInputItem],
+    ) -> None:
+        """A soft nudge, never a stop: at the turn a skill asks for, then every few turns."""
+        after = context.context.get(WRAP_UP_AFTER_TURNS_KEY)
+        requests = getattr(getattr(context, "usage", None), "requests", None)
+        if not isinstance(after, int) or not isinstance(requests, int):
+            return
+        turns_used = requests + 1
+        if turns_used < after or (turns_used - after) % _WRAP_UP_REPEAT_TURNS:
+            return
+        content = (
+            f"[NOTICE] You have used {turns_used} turns. Wrap up soon: take only the steps "
+            "you still need to answer your task, then call agent_finish with what you have, "
+            "listing anything unresolved in open_items instead of tracing it further."
+        )
+        input_items.append({"role": "user", "content": content})
+
     def _maybe_warn_budget(
         self,
         context: RunContextWrapper[dict[str, Any]],
diff --git a/strix/skills/__init__.py b/strix/skills/__init__.py
index 1e66fae0b4..fd53c60c94 100644
--- a/strix/skills/__init__.py
+++ b/strix/skills/__init__.py
@@ -270,6 +270,19 @@ def _candidate_skill_files(skill_name: str) -> list[Path]:
     return _bare_skill_files(skill_name)
 
 
+def skill_wrap_up_after_turns(skill_names: list[str]) -> int | None:
+    """Smallest ``wrap_up_after_turns`` set in these skills' frontmatter, if any."""
+    turns = []
+    for skill_name in skill_names:
+        candidates = _candidate_skill_files(skill_name)
+        if len(candidates) != 1:
+            continue
+        value = _read_skill_metadata(candidates[0]).get("wrap_up_after_turns", "")
+        if value.isdigit() and int(value) > 0:
+            turns.append(int(value))
+    return min(turns, default=None)
+
+
 def load_skills(skill_names: list[str], *, supports_images: bool = True) -> dict[str, str]:
     """Load skill markdown bodies (frontmatter stripped) by name.