diff --git a/doc/plyngent.example.toml b/doc/plyngent.example.toml
index bd4e16a..3125034 100644
--- a/doc/plyngent.example.toml
+++ b/doc/plyngent.example.toml
@@ -41,6 +41,9 @@ max_context_tokens = 200000
 # developer (default) | user | synthetic_tool | none
 # (legacy system is treated as developer)
 # todo_nag_strategy = "developer"
+# Append-only developer playbook checkpoints every N prompt_tokens (0 = off).
+# directive_reminder_tokens = 100000
+# directive_reminder_text = ""
 
 # Optional: custom compact/summarisation prompts (empty = built-in defaults).
 # compact_system_prompt = ""
diff --git a/src/plyngent/cli/slash.py b/src/plyngent/cli/slash.py
index 634fb51..ba4d56a 100644
--- a/src/plyngent/cli/slash.py
+++ b/src/plyngent/cli/slash.py
@@ -544,6 +544,9 @@ def status_cmd(state: ReplState) -> None:
         f"context_tokens={ctx_tilde}{ctx_tokens}/{ctx_budget} ({ctx_tag})  "
         f"context_chars={ctx_chars}  "
         f"tool_result_max={state.agent.max_tool_result_chars}\n"
+        f"peak_prompt_tokens={state.agent.peak_prompt_tokens}  "
+        f"reminder_band={state.agent.reminder_last_band}  "
+        f"reminder_every={state.agent.directive_reminder_tokens or 'off'}\n"
         f"last_request={last_req.format_line()}\n"
         f"usage_last_turn={last_u.format_line(billed=True)}  "
         f"rounds={last_rounds}\n"
diff --git a/src/plyngent/cli/state.py b/src/plyngent/cli/state.py
index 9bb0e4a..ec87d8f 100644
--- a/src/plyngent/cli/state.py
+++ b/src/plyngent/cli/state.py
@@ -236,7 +236,14 @@ class ReplState:
             agent_cfg.tool_directives,
         )
         on_limit = prompt_continue_limit_async if self.interactive_limits else None
-        return ChatAgent(
+        peak = 0
+        band = 0
+        last_req = None
+        if hasattr(self, "agent"):
+            peak = self.agent.peak_prompt_tokens
+            band = self.agent.reminder_last_band
+            last_req = self.agent.last_request_usage
+        agent = ChatAgent(
             self.client,
             model=self.model,
             tools=self._tool_registry(),
@@ -251,7 +258,14 @@ class ReplState:
             max_context_tokens=agent_cfg.max_context_tokens,
             todo_stack=self.todo_stack,
             todo_nag_strategy=agent_cfg.todo_nag_strategy,
+            directive_reminder_tokens=agent_cfg.directive_reminder_tokens,
+            directive_reminder_text=agent_cfg.directive_reminder_text or None,
+            peak_prompt_tokens=peak,
+            reminder_last_band=band,
         )
+        if last_req is not None and not last_req.is_zero():
+            agent.last_request_usage = last_req
+        return agent
 
     def rebuild_client(self) -> None:
         """Recreate client and agent after provider/model/tools change."""
diff --git a/src/plyngent/config/models.py b/src/plyngent/config/models.py
index 8315d73..2529db5 100644
--- a/src/plyngent/config/models.py
+++ b/src/plyngent/config/models.py
@@ -99,6 +99,12 @@ class AgentConfig(Struct, omit_defaults=True):
     # developer | user | synthetic_tool | none  (legacy "system" → developer)
     todo_nag_strategy: str = "developer"
 
+    # Append-only developer playbook checkpoints when last prompt_tokens crosses
+    # N * this interval (0 = off). See agent/directive_checkpoint.py.
+    directive_reminder_tokens: int = 100_000
+    # Optional short checklist body (empty = built-in hard-constraint list).
+    directive_reminder_text: str = ""
+
     # Compact / summarisation prompts (empty = use built-in defaults).
     compact_system_prompt: str = ""
     compact_user_prefix: str = ""