feat(evolution): give review agent full context, add knowledge signal, polish UX

agent/evolution/config.py

@@ -14,7 +14,7 @@ from typing import Any
 # until release; enable via ``self_evolution_enabled``.
 DEFAULT_ENABLED = False
 DEFAULT_IDLE_MINUTES = 15
-DEFAULT_MIN_TURNS = 6
+DEFAULT_MIN_TURNS = 8
 # Max review steps for the isolated evolution agent. Kept small (not exposed as
 # config): the review is meant to be cheap and focused, not a long autonomous run.
 DEFAULT_MAX_STEPS = 12

agent/evolution/executor.py

@@ -175,6 +175,9 @@ _WATCH_SUBDIRS = ("MEMORY.md", "skills", "knowledge", "output")
 # Subpaths under memory/ to ignore: evolution's own bookkeeping + the nightly
 # dream diary, none of which count as a user-facing change signal.
 _MEMORY_IGNORE = (".evolution_backups", "dreams", "evolution")
+# Files the skill subsystem maintains automatically (the enable/disable index).
+# Not an evolution result, so a rewrite must not count as a change signal.
+_WATCH_IGNORE_NAMES = ("skills_config.json",)
 
 
 def _workspace_snapshot(workspace_dir) -> dict:
@@ -195,6 +198,8 @@ def _workspace_snapshot(workspace_dir) -> dict:
         for p in root.rglob("*"):
             if not p.is_file():
                 continue
+            if p.name in _WATCH_IGNORE_NAMES:
+                continue
             try:
                 st = p.stat()
                 snap[str(p.relative_to(ws))] = (st.st_mtime, st.st_size)
@@ -269,10 +274,8 @@ def run_evolution_for_session(
         new_messages = all_messages[done:]
         transcript = _build_transcript(new_messages)
         if not transcript.strip():
-            # Routine no-op: the per-minute scan hits every idle session, so keep
-            # this at debug to avoid spamming the log.
-            logger.debug(f"[Evolution] session={session_id}: no new messages, skip")
-            # Advance the cursor anyway so we don't re-scan the same tail.
+            # Routine no-op: the per-minute scan hits every idle session. Advance
+            # the cursor so we don't re-scan the same tail; no log (pure noise).
             agent._evo_done_msg_count = total_msgs
             return False
 
@@ -334,17 +337,23 @@ def run_evolution_for_session(
             str(workspace_dir),
         )
         review_agent = agent_bridge.create_agent(
-            system_prompt=EVOLUTION_SYSTEM_PROMPT,
+            system_prompt="",
             tools=review_tools,
             description="Self-evolution review agent",
             max_steps=cfg.max_steps,
             workspace_dir=str(workspace_dir),
             skill_manager=getattr(agent, "skill_manager", None),
             memory_manager=getattr(agent, "memory_manager", None),
-            enable_skills=False,
+            enable_skills=True,
+            runtime_info=getattr(agent, "runtime_info", None),
         )
         # Reuse the live model so it follows the user's configured model.
         review_agent.model = agent.model
+        # Inject the evolution task brief AFTER the full system prompt: the agent
+        # gets the full context (tools, workspace, user preferences, memory, time)
+        # AND its evolution-specific instructions on top, instead of one
+        # overwriting the other.
+        review_agent.extra_system_suffix = EVOLUTION_SYSTEM_PROMPT
 
         logger.info(
             f"[Evolution] backup {backup_id} ({_backup_n} files) → running review agent"

agent/evolution/prompts.py

@@ -7,7 +7,7 @@ language (instructed at the end of the prompt).
 
 Design goals (see ref/hermes-agent background_review for inspiration):
   - Default to doing NOTHING. Evolution is the exception, not the rule.
-  - Three signal types: memory, skill, unfinished task.
+  - Signal types: skill, unfinished task, memory, knowledge.
   - An explicit "do NOT capture" list to avoid self-poisoning over time.
   - Generic examples only — never bake in domain-specific business terms.
 """
@@ -72,12 +72,11 @@ them. When their signal is clear, act; do not be shy here.
    reply/decision, do NOTHING and stay [SILENT] — do not nag or ping the user.
    You only ever notify the user as a side effect of having actually done work.
 
-3. MEMORY — LAST resort, and you are only a SAFETY NET here, not the primary
-   writer. The main assistant already writes memory DURING the conversation, and
-   a nightly pass consolidates daily notes into long-term memory. Prefer fixing
-   a skill (above) over writing memory whenever the fact belongs in a skill.
-   Act ONLY on something the main assistant clearly MISSED that does not belong
-   in any skill.
+3. MEMORY — RARE, last resort. Default to writing NOTHING here. The main
+   assistant already writes memory during the chat, and a nightly pass plus
+   context-overflow saves are dedicated safety nets — so memory is almost always
+   already covered without you. Skip unless the main assistant clearly missed a
+   durable fact that belongs in no skill AND would visibly change future replies.
    - MEMORY.md is the curated long-term index, auto-loaded into EVERY future
      conversation. Treat it as precious: edit it in place to CORRECT a wrong
      fact, or append a new durable preference/decision/lesson — but do so
@@ -92,6 +91,12 @@ them. When their signal is clear, act; do not be shy here.
    - If it is already captured anywhere (check MEMORY.md AND the daily file
      first), do NOTHING.
 
+4. KNOWLEDGE — only if the conversation produced durable, reusable reference
+   knowledge on a topic (the kind worth looking up again) that the main
+   assistant did NOT already save to `knowledge/`. Add or update the relevant
+   file there. Like memory, this is the exception: skip routine Q&A, and if the
+   topic is already covered in `knowledge/`, do NOTHING rather than duplicate.
+
 # Do NOT capture (these poison future behavior)
 
 - Environment failures: missing binaries, unset credentials, uninstalled
@@ -140,9 +145,6 @@ def build_review_user_message(transcript: str, protected_skills: list = None) ->
     ``protected_skills`` lists skill names that must never be edited (built-in
     skills shipped with the product). Surfaced so the agent avoids them.
     """
-    from datetime import datetime
-    today = datetime.now().strftime("%Y-%m-%d")
-
     protected_note = ""
     if protected_skills:
         names = ", ".join(sorted(protected_skills))
@@ -152,12 +154,10 @@ def build_review_user_message(transcript: str, protected_skills: list = None) ->
         )
     return (
         "Here is the conversation transcript that just went idle. Review it per "
-        "your instructions and act on any clear signal. Prefer fixing a skill at "
-        "its source over writing memory whenever the fact belongs in a skill.\n"
-        f"Today is {today}. Only if a fact genuinely belongs in memory (and not "
-        f"in a skill): append one short bullet to the daily file "
-        f"`memory/{today}.md` for a new fact, or edit MEMORY.md in place to "
-        f"correct an existing wrong fact."
+        "your instructions. Acting is the exception: the main value is fixing or "
+        "creating a skill and finishing promised work. Memory and knowledge are "
+        "rare last resorts — stay [SILENT] unless there is a clear, durable signal "
+        "not already covered."
         f"{protected_note}\n"
         "<transcript>\n"
         f"{transcript}\n"