A lightweight monitoring system that provides 3-5 word real-time status updates for active sub-agents to keep workflows transparent and efficient.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-summary
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agent-summary using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Agent Summary?
The Agent Summary skill implements a progress tracking mechanism inspired by the Claude Code AgentSummary system. It is designed to provide developers and automated systems with high-level visibility into the activities of sub-agents without requiring manual log inspection. By integrating this into your Openclaw Skills environment, you can maintain a clear overview of complex, multi-threaded agent operations while minimizing cognitive overhead.
This skill periodically analyzes sub-agent session histories to generate concise summaries of their current tasks. It prioritizes technical accuracy, replacing vague descriptions like "investigating" with specific actions such as "Reading runAgent.ts." By reusing parent agent context, it ensures that monitoring remains cost-effective and performant through high cache hit rates.
Agent Summary Use Cases
- Tracking the progress of long-running background tasks across multiple sub-agents in a single view.
- Providing human-in-the-loop oversight with clear, 3-5 word status updates for each active worker.
- Identifying bottlenecks or stalling agents by seeing exactly which file or function is currently being processed.
- Automating status reporting within complex Openclaw Skills orchestration pipelines.
How Agent Summary Works
- The system identifies all currently active workers using the subagents list command.
- It retrieves the most recent session history (typically the last 2 messages) for each active sub-agent.
- A lightweight query is executed to distill the current activity into a specific 3-5 word summary.
- The system compares the new summary with the previous state to determine if an update is necessary.
- If a change is detected or an agent finishes, a formatted status tree is reported to the user or coordinator.
Agent Summary Setup
To deploy Agent Summary within your Openclaw Skills configuration, ensure you have the necessary subagent and session history permissions.
# Check active workers to confirm the environment is ready
openclaw subagents list
You can then integrate the logic into your HEARTBEAT.md or a custom coordinator script to poll for updates every 30 seconds.
Agent Summary Data Schema & Taxonomy
| Component | Description | Format |
|---|---|---|
| Agent ID | The unique identifier for the worker agent | String |
| Status Summary | A 3-5 word description of the current task | String |
| Lifecycle State | Current state: Active, Completed, or Failed | Enum |
| Quiet Hours | Period where non-critical alerts are silenced | 23:00-08:00 |
Agent Summary Advanced Features
- Context reuse strategy that minimizes token expenditure while maximizing response speed.
- Intelligent notification filtering that only triggers alerts when a sub-agent changes its specific task or reaches a terminal state.
- Seamless integration with HEARTBEAT.md for automated, autonomous background monitoring.
- Built-in "Quiet Hours" logic to prevent non-essential notifications during late-night hours.
- Support for multi-agent coordination, allowing a central Openclaw Skills controller to manage a fleet of specialized workers.
SKILL.md
Loading
Related Openclaw Skills
Web Search to Feishu Bitable
Automatically search keywords, extract web content, and organize data into a structured Feishu Bitable personal knowledge base.
Agent Summary System
A lightweight progress monitoring system that generates 3-5 word status updates for active sub-agents every 30 seconds.
Lobster Agentic Engineering
A sophisticated framework for agent-led engineering that prioritizes eval-first execution, task decomposition, and cost-efficient model routing.
Agent Link
Agent Link is a secure communication framework that enables OpenClaw instances and agents to exchange messages across different computers via a dedicated relay server.
PromptBuddy
PromptBuddy is an intelligent middleware skill that automatically reformats and optimizes user queries into structured prompts for superior AI reasoning.
Attachment Inject
A high-efficiency context management skill that injects dynamic data into AI sessions as attachments rather than system prompt modifications.
Featured*
