feat(ai): support stdin for Gemini CLI and verify with integration test

tests/mock_gemini_cli.py

@@ -0,0 +1,83 @@
+import sys
+import json
+import subprocess
+import os
+
+def main():
+    # The GUI calls: <binary> run --output-format stream-json
+    # The prompt is now passed via stdin.
+    
+    # Debug log to stderr
+    sys.stderr.write(f"DEBUG: mock_gemini_cli called with args: {sys.argv}\n")
+    
+    # Read prompt from stdin for debug
+    prompt = sys.stdin.read()
+    sys.stderr.write(f"DEBUG: Received prompt via stdin ({len(prompt)} chars)\n")
+    sys.stderr.flush()
+
+    if "run" not in sys.argv:
+        return
+
+    # Simulate the 'BeforeTool' hook by calling the bridge directly.
+    bridge_path = os.path.abspath("scripts/cli_tool_bridge.py")
+    
+    tool_call = {
+        "tool_name": "read_file",
+        "tool_input": {"path": "test.txt"}
+    }
+    
+    sys.stderr.write(f"DEBUG: Calling bridge at {bridge_path}\n")
+    sys.stderr.flush()
+
+    # Bridge reads from stdin
+    process = subprocess.Popen(
+        [sys.executable, bridge_path],
+        stdin=subprocess.PIPE,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.PIPE,
+        text=True
+    )
+    stdout, stderr = process.communicate(input=json.dumps(tool_call))
+    
+    sys.stderr.write(f"DEBUG: Bridge stdout: {stdout}\n")
+    sys.stderr.write(f"DEBUG: Bridge stderr: {stderr}\n")
+    sys.stderr.flush()
+
+    try:
+        decision_data = json.loads(stdout.strip())
+        decision = decision_data.get("decision")
+    except Exception as e:
+        sys.stderr.write(f"DEBUG: Failed to parse bridge output: {e}\n")
+        decision = "deny"
+
+    # Output JSONL to stdout
+    if decision == "allow":
+        print(json.dumps({
+            "type": "tool_use", 
+            "name": "read_file", 
+            "args": {"path": "test.txt"}
+        }), flush=True)
+        
+        print(json.dumps({
+            "type": "message", 
+            "text": "I read the file. It contains: 'Hello from mock!'"
+        }), flush=True)
+        
+        print(json.dumps({
+            "type": "result", 
+            "usage": {"total_tokens": 50},
+            "session_id": "mock-session-123"
+        }), flush=True)
+    else:
+        print(json.dumps({
+            "type": "message", 
+            "text": f"Tool execution was denied. Decision: {decision}"
+        }), flush=True)
+        print(json.dumps({
+            "type": "result", 
+            "usage": {"total_tokens": 10},
+            "session_id": "mock-session-denied"
+        }), flush=True)
+
+if __name__ == "__main__":
+    main()