Harness: runner, generator, tasks T01 T13 T14 T15, CLAUDE.md

harness/proxy.py

@@ -0,0 +1,80 @@
+"""Tool proxy between agent and MCP: whitelist, budget, prefix filter, trajectory log."""
+import json
+import re
+import time
+
+MODEL_TOOLS = {
+    "sap_search_object", "sap_pull_source", "sap_object_structure", "sap_object_members",
+    "sap_usage_references", "sap_element_info", "sap_inactive_objects", "sap_short_dumps",
+    "sap_sql_query", "sap_create_object", "sap_push_source", "sap_push_element",
+    "sap_push_message", "sap_activate", "sap_syntax_check", "sap_check_object",
+    "sap_run_unit_test", "sap_atc_run", "sap_pretty_print", "sap_run_class",
+}
+WRITE_TOOLS = {"sap_create_object", "sap_push_source", "sap_push_element",
+               "sap_push_message", "sap_activate"}
+RUN_PREFIX = re.compile(r"^Z\d[0-9A-Z]{5,6}_", re.I)
+
+
+class BudgetExceeded(Exception):
+    pass
+
+
+class ToolProxy:
+    def __init__(self, mcp, prefix, budget, log_path):
+        self.mcp = mcp
+        self.prefix = prefix.upper()
+        self.max_calls = budget.get("max_tool_calls", 60)
+        self.max_activations = budget.get("max_activations", 15)
+        self.calls = 0
+        self.activations = 0
+        self.fail_streak = 0
+        self.max_fail_streak = 0
+        self.log = open(log_path, "a")
+
+    def schemas(self):
+        return [t for t in self.mcp.list_tools() if t["name"] in MODEL_TOOLS]
+
+    def _foreign(self, name):
+        n = (name or "").upper()
+        return bool(RUN_PREFIX.match(n)) and not n.startswith(self.prefix)
+
+    def _filter(self, tool, text):
+        if tool not in ("sap_search_object", "sap_usage_references"):
+            return text
+        try:
+            data = json.loads(text)
+        except ValueError:
+            return text
+        if isinstance(data, list):
+            data = [d for d in data if not self._foreign(d.get("name") if isinstance(d, dict) else "")]
+            return json.dumps(data)
+        return text
+
+    def call(self, tool, args):
+        entry = {"t": time.time(), "tool": tool, "args": args}
+        if tool not in MODEL_TOOLS:
+            result = (True, f"Tool {tool} is not available.")
+        elif self.calls >= self.max_calls:
+            raise BudgetExceeded(f"max_tool_calls={self.max_calls}")
+        else:
+            self.calls += 1
+            name = str(args.get("objectName", "")).upper()
+            if tool in WRITE_TOOLS and self._foreign(name):
+                result = (True, f"{name} is not available.")
+            else:
+                if tool in WRITE_TOOLS and (tool == "sap_activate" or args.get("activate", True)):
+                    self.activations += 1
+                err, text = self.mcp.call(tool, args)
+                result = (err, self._filter(tool, text))
+                if tool in WRITE_TOOLS:
+                    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)
+        entry["is_error"], entry["result"] = result[0], result[1][:20000]
+        self.log.write(json.dumps(entry) + "\n")
+        self.log.flush()
+        return result
+
+    def note(self, kind, content):
+        self.log.write(json.dumps({"t": time.time(), "kind": kind, "content": content}) + "\n")
+        self.log.flush()