Loop guard (3 identical pushes), end_reason and activation error records per run; thinking off for the stage 1 baseline

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014aUaQeLnwbb1zTpN7kHeat

harness/proxy.py

@@ -1,4 +1,5 @@
 """Tool proxy between agent and MCP: whitelist, budget, prefix filter, trajectory log."""
+import hashlib
 import json
 import re
 import time
@@ -30,6 +31,23 @@ VARIANTS = {
                  "packageName": "package", "includeType": "include", "className": "class_name"},
     },
 }
+
+
+def activation_messages(text, is_error=False):
+    """Error messages (severity E) of a write result; the whole text if the call itself failed."""
+    try:
+        obj = json.loads(text)
+    except (TypeError, ValueError):
+        return [str(text)[:300]] if is_error else []
+    if not isinstance(obj, dict):
+        return []
+    msgs = (obj.get("activation") or {}).get("messages") or obj.get("messages") or []
+    out = [str(m.get("message", ""))[:300] for m in msgs if isinstance(m, dict) and m.get("severity") == "E"]
+    if not out and is_error:
+        out = [str(obj.get("message") or text)[:300]]
+    return out
+
+
 RUN_PREFIX = re.compile(r"^Z\d[0-9A-Z]{5,6}_", re.I)
 
 
@@ -49,6 +67,10 @@ class ToolProxy:
         self.activations = 0
         self.fail_streak = 0
         self.max_fail_streak = 0
+        self.activation_failures = 0  # write calls that failed
+        self.activation_errors = []  # unique error messages of failed writes
+        self.same_push_streak = 0  # identical sap_push_source (object + source hash) in a row
+        self._last_push = None
         self.log = open(log_path, "a")
         self.adt = None
         self.fallbacks = []  # writes that needed the ADT activation fallback
@@ -125,6 +147,11 @@ class ToolProxy:
             else:
                 if tool in WRITE_TOOLS and (tool == "sap_activate" or args.get("activate", True)):
                     self.activations += 1
+                if tool == "sap_push_source":
+                    key = (name, str(args.get("includeType", "")),
+                           hashlib.md5(str(args.get("source", "")).encode()).hexdigest())
+                    self.same_push_streak = self.same_push_streak + 1 if key == self._last_push else 1
+                    self._last_push = key
                 err, text = self.mcp.call(tool, args)
                 if tool == "sap_push_source":
                     if str(args.get("objectType", "")).upper() in ("PROG", "FUNC") and not args.get("includeType"):
@@ -139,6 +166,11 @@ class ToolProxy:
                     failed = err or '"success":false' in text.replace(" ", "")
                     self.fail_streak = self.fail_streak + 1 if failed else 0
                     self.max_fail_streak = max(self.max_fail_streak, self.fail_streak)
+                    if failed:
+                        self.activation_failures += 1
+                        for m in activation_messages(text, err):
+                            if m not in self.activation_errors and len(self.activation_errors) < 30:
+                                self.activation_errors.append(m)
         entry["is_error"], entry["result"] = result[0], result[1][:20000]
         self.log.write(json.dumps(entry) + "\n")
         self.log.flush()