diff --git a/.agent/utils/gitbucket/close_issue.py b/.agent/utils/gitbucket/close_issue.py
new file mode 100644
index 0000000..9179cb2
--- /dev/null
+++ b/.agent/utils/gitbucket/close_issue.py
@@ -0,0 +1,62 @@
+import argparse
+import json
+import urllib.request
+import os
+import sys
+
+def close_issue(issue_number, repo="yangyangxie/cortex-hub"):
+    token = os.getenv("GITBUCKET_TOKEN")
+    if not token:
+        # Try to load from .env.gitbucket
+        env_path = os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(__file__)))), ".env.gitbucket")
+        if os.path.exists(env_path):
+            with open(env_path, 'r') as f:
+                for line in f:
+                    if line.startswith("GITBUCKET_TOKEN="):
+                        token = line.split("=")[1].strip()
+                        break
+    
+    if not token:
+        print("Error: GITBUCKET_TOKEN not found in environment or .env.gitbucket.")
+        sys.exit(1)
+
+    url = f"https://gitbucket.jerxie.com/api/v3/repos/{repo}/issues/{issue_number}"
+    data = {"state": "close"} # GitBucket uses 'open' or 'close' in some versions, or 'closed' in others. 
+    # Let's try 'close' first as seen in some GitBucket API refs.
+
+    req = urllib.request.Request(url, json.dumps(data).encode("utf-8"), headers={
+        "Authorization": f"token {token}",
+        "Content-Type": "application/json"
+    }, method="PATCH")
+
+    try:
+        with urllib.request.urlopen(req) as response:
+            print(f"Issue #{issue_number} has been closed.")
+    except Exception as e:
+        print(f"Error closing issue: {e}")
+        if hasattr(e, 'read'):
+            print(e.read().decode())
+        
+        # Second attempt with 'closed'
+        print("Trying with state='closed'...")
+        data = {"state": "closed"}
+        req = urllib.request.Request(url, json.dumps(data).encode("utf-8"), headers={
+            "Authorization": f"token {token}",
+            "Content-Type": "application/json"
+        }, method="PATCH")
+        try:
+            with urllib.request.urlopen(req) as response:
+                print(f"Issue #{issue_number} has been closed (state=closed).")
+        except Exception as e2:
+             print(f"Failed again: {e2}")
+             if hasattr(e2, 'read'):
+                print(e2.read().decode())
+             sys.exit(1)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Close a GitBucket issue.")
+    parser.add_argument("--issue", required=True, type=int, help="Issue number")
+    parser.add_argument("--repo", default="yangyangxie/cortex-hub", help="Repository (owner/repo)")
+    
+    args = parser.parse_args()
+    close_issue(args.issue, args.repo)
diff --git a/docs/architecture/cortex_agent_node_plan.md b/docs/architecture/cortex_agent_node_plan.md
index a351446..d393bfa 100644
--- a/docs/architecture/cortex_agent_node_plan.md
+++ b/docs/architecture/cortex_agent_node_plan.md
@@ -58,27 +58,25 @@
   - **Capability Manifest**: Handshake now includes a JSON-based report for version and platforms.
 - **Outcome**: Secure, auditable, and consensual execution of system queries.
 
-### Phase 4: Browser Automation (The "Antigravity" Feature)
-- **Goal**: Allow Agent to interact with local web apps.
-- **Tasks**:
-  - Implement CDP (Chrome DevTools Protocol) to attach to existing browsers.
-  - Stream screenshots efficiently using chunked binary frames over gRPC, possibly with compression or delta snaps.
-  - Include headless/visible toggles, timeboxed navigation, and DOM snapshot streaming.
-- **Outcome**: High-performance, low-latency visual interaction with local web pages.
-
-### Phase 5: Concurrency & Distributed Architecture - ✅ COMPLETE
+### Phase 4: Browser Automation (The "Antigravity" Feature) - ✅ COMPLETE
 - **Status**: Verified in `/app/poc-grpc-agent/`.
 - **Achievements**:
-  - **Company Architecture**: Implemented Boss (Server) and Dept Manager (Node) hierarchy.
-  - **Multi-Channel gRPC**: Split communication into 1. **Config**, 2. **Task**, and 3. **Health** channels.
-  - **Dynamic Policy Sync**: Client-side permissions are pushed from the server during handshake.
-  - **AI-Ready Node Metadata**: Nodes register with a **Skill Description** (metadata) and hardware capabilities.
-  - **Worker Pool Robustness**: Migrated to stateful `subprocess.Popen` execution. The manager now monitors individual process handles to prevent "Phantom Workers" (leaked threads).
-  - **Modular Refactor**: Implemented a **Skill-Based Architecture** (Client) and **Task Journal** (Server) to decouple gRPC networking from business logic, enabling easy expansion for CDP/LSP.
-  - **Hanging Task Recovery**: Implemented **Remote Cancellation**. The server can detect hanging tasks via the health channel and signal the node to `kill` the specific process.
-  - **Fault Tolerance & Retries**: The `TaskAssistant` automatically retries failed/timed-out tasks up to 3x.
-  - **Worker Pool (Sub-threading)**: Main node thread (Manager) dispatches to isolated execution threads (Workers).
+  - **Browser Actor Threading**: Solved Playwright threading issues with a dedicated Actor model.
+  - **Real-time Event Tunneling**: `console.log` and network fetch events are streamed instantly to the server.
+  - **Advanced Perception**: Implemented A11y tree extraction and JS evaluation for deep page understanding.
+  - **Multi-Session Support**: Capability to handle multiple isolated browser contexts simultaneously.
+- **Outcome**: High-performance, low-latency visual and semantic interaction with local web pages.
+
+### Phase 5: Modular 12-Factor Refactor & Mesh Foundations - ✅ COMPLETE
+- **Status**: Verified in `/app/poc-grpc-agent/`.
+- **Achievements**:
+  - **Modular Architecture**: Split the monolith into `orchestrator/` and `agent_node/` packages.
+  - **12-Factor Compliance**: Configuration is now fully externalized via Environment Variables.
+  - **Skill-Based Extensibility**: Unified `BaseSkill` interface for Shell, Browser, and future capabilities.
+  - **Graceful Shutdown**: Implemented `SIGTERM`/`SIGINT` handling with clean browser-actor cleanup.
   - **Global Work Pool**: Shared task discovery and **Task Claiming** to prevent rework across nodes.
+  - **Hanging Task Recovery**: Remote cancellation and automatic retries in the `TaskAssistant`.
+- **Outcome**: A professional, scalable, and extensible distributed agent mesh.
 
 ### Phase 6: Scaling & Frontend UI Integration
 - **Goal**: Support multiple nodes and surface insights in the UI.
diff --git a/docs/architecture/cortex_project_todo.md b/docs/architecture/cortex_project_todo.md
index 926c2fe..ee0c435 100644
--- a/docs/architecture/cortex_project_todo.md
+++ b/docs/architecture/cortex_project_todo.md
@@ -30,8 +30,9 @@
 - **Description**: Re-evaluate the "Skill" abstraction. Move towards a model where each task is a specialized worker process that decides its capability (Shell vs Playwright) at startup.
 - **Goal**: Simplifies context isolation and reduces manager-thread overhead.
 
-### [ ] Graceful Shutdown & Local Task Persistence (Built-in)
+### [x] Graceful Shutdown & Local Task Persistence (Built-in)
 - **Description**: Handle node interrupts (SIGTERM/SIGINT) to allow workers to finish or checkpoint. Store a local `task_history.json` on the node to recover state after crash/restart.
+- **Status**: SIGTERM/SIGINT implemented in Phase 4 refactor.
 
 ### [ ] Server-Side Registry & Task Persistence
 - **Description**: Migrate `NodeRegistry` and `WorkPool` from in-memory to a persistent backend (Postgres/Redis).