[kimi] Add JSON validation guard to queue.json writes (#952) (#995)

scripts/loop_guard.py

@@ -202,7 +202,11 @@ def load_queue() -> list[dict]:
                 # Persist the cleaned queue so stale entries don't recur
                 _save_cleaned_queue(data, open_numbers)
         return ready
-    except (json.JSONDecodeError, OSError):
+    except json.JSONDecodeError as exc:
+        print(f"[loop-guard] WARNING: Corrupt queue.json ({exc}) — returning empty queue")
+        return []
+    except OSError as exc:
+        print(f"[loop-guard] WARNING: Cannot read queue.json ({exc}) — returning empty queue")
         return []
 
 

scripts/triage_score.py

@@ -41,6 +41,7 @@ REPO_SLUG = os.environ.get("REPO_SLUG", "rockachopa/Timmy-time-dashboard")
 TOKEN_FILE = Path.home() / ".hermes" / "gitea_token"
 REPO_ROOT = Path(__file__).resolve().parent.parent
 QUEUE_FILE = REPO_ROOT / ".loop" / "queue.json"
+QUEUE_BACKUP_FILE = REPO_ROOT / ".loop" / "queue.json.bak"
 RETRO_FILE = REPO_ROOT / ".loop" / "retro" / "triage.jsonl"
 QUARANTINE_FILE = REPO_ROOT / ".loop" / "quarantine.json"
 CYCLE_RETRO_FILE = REPO_ROOT / ".loop" / "retro" / "cycles.jsonl"
@@ -342,9 +343,38 @@ def run_triage() -> list[dict]:
     ready = [s for s in scored if s["ready"]]
     not_ready = [s for s in scored if not s["ready"]]
 
+    # Save backup before writing (if current file exists and is valid)
+    if QUEUE_FILE.exists():
+        try:
+            json.loads(QUEUE_FILE.read_text())  # Validate current file
+            QUEUE_BACKUP_FILE.write_text(QUEUE_FILE.read_text())
+        except (json.JSONDecodeError, OSError):
+            pass  # Current file is corrupt, don't overwrite backup
+
+    # Write new queue file
     QUEUE_FILE.parent.mkdir(parents=True, exist_ok=True)
     QUEUE_FILE.write_text(json.dumps(ready, indent=2) + "\n")
 
+    # Validate the write by re-reading and parsing
+    try:
+        json.loads(QUEUE_FILE.read_text())
+    except (json.JSONDecodeError, OSError) as exc:
+        print(f"[triage] ERROR: queue.json validation failed: {exc}", file=sys.stderr)
+        # Restore from backup if available
+        if QUEUE_BACKUP_FILE.exists():
+            try:
+                backup_data = QUEUE_BACKUP_FILE.read_text()
+                json.loads(backup_data)  # Validate backup
+                QUEUE_FILE.write_text(backup_data)
+                print(f"[triage] Restored queue.json from backup")
+            except (json.JSONDecodeError, OSError) as restore_exc:
+                print(f"[triage] ERROR: Backup restore failed: {restore_exc}", file=sys.stderr)
+                # Write empty list as last resort
+                QUEUE_FILE.write_text("[]\n")
+        else:
+            # No backup, write empty list
+            QUEUE_FILE.write_text("[]\n")
+
     # Write retro entry
     retro_entry = {
         "timestamp": datetime.now(timezone.utc).isoformat(),