Does this skill store my agent data in a database?

No, the observability skill uses an ephemeral, in-memory architecture that streams events directly from your local filesystem and clears upon restart.

Can I use this to debug agent loops?

Yes, the real-time trace shows exactly which tools are being called in sequence, making it easy to identify if an agent is stuck in a repetitive cycle.

How do I access the dashboard once it's started?

Once you trigger 'start observability', the dashboard UI is accessible via your web browser at http://localhost:5172.

Which ports does the dashboard use?

The dashboard uses port 5172 for the client UI and port 4000 for the backend server and WebSocket stream.

What local files does this skill monitor?

It tails the JSONL event files located in your ~/.claude/history/raw-outputs/ directory to capture multi-agent activity.

Agent Observability Dashboard

Name: Agent Observability Dashboard
Author: blackpwnguin

byblackpwnguin

分析与监控

Monitors multi-agent activity in real-time through a specialized dashboard, providing live insights into tool calls and workflow traces.

关于

The Agent Observability Dashboard provides a comprehensive real-time view of multi-agent interactions within the PAI ecosystem. By leveraging WebSocket streaming and filesystem monitoring, it allows developers to visualize agent sessions, track tool call execution, and debug complex multi-agent workflows as they happen. It is an essential tool for identifying bottlenecks, loops, or errors in agentic chains by transforming raw JSONL logs into an intuitive, live-updating visual interface.

主要功能

Real-time WebSocket streaming of agent activity
Live tracking of tool calls and hook execution
Visual session timelines and execution traces
Automated filesystem watching for event log updates
Simple CLI management for starting and stopping the service
0 GitHub stars

使用场景

Debugging stuck or looping multi-agent workflows
Monitoring tool call sequences during complex autonomous tasks
Visualizing agent communication and session lifecycles

关于

主要功能

Real-time WebSocket streaming of agent activity
Live tracking of tool calls and hook execution
Visual session timelines and execution traces
Automated filesystem watching for event log updates
Simple CLI management for starting and stopping the service
0 GitHub stars

使用场景

Debugging stuck or looping multi-agent workflows
Monitoring tool call sequences during complex autonomous tasks
Visualizing agent communication and session lifecycles