A standardized protocol for real-time AI agent status checks, providing instant visibility into load, interruptibility, and task readiness.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install goat-agent-pulse
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 goat-agent-pulse 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 Pulse?
Agent Pulse is a specialized status-check protocol designed for AI agents to communicate their current operational state without the overhead of heavy reasoning. By utilizing a rule-based approach, it provides a consistent, low-latency snapshot of an agent's workload and availability, ensuring that users and orchestrators know exactly when it is appropriate to interrupt or assign new tasks.
As a core component within the Openclaw Skills ecosystem, Agent Pulse prioritizes baseline status over self-influenced metrics. This ensures that the act of querying the agent's status does not artificially inflate its perceived busy-ness. It is the ideal solution for maintaining high coordination efficiency in multi-agent environments or complex human-AI workflows.
Agent Pulse Use Cases
- Determining if an agent is currently available for a high-priority task insertion.
- Checking if an agent is blocked by external dependencies or waiting for user input.
- Standardizing status reporting across a fleet of different AI agents.
- Providing quick, non-conversational feedback for automated system health monitors.
- Reducing token usage by substituting complex 'Are you busy?' queries with a fixed-format pulse check.
How Agent Pulse Works
- The skill is triggered by specific commands like /pulse or natural language queries regarding the agent's availability.
- The system gathers low-cost signals such as active project advancement, queued messages, and recent tool execution activity.
- Collected signals are processed through the pulse_eval.py script to map raw data to standardized status levels.
- The agent generates a fixed-contract output card containing four specific fields: status, interruptibility, acceptNewTask, and a short reason.
- The result is returned as a compact text block, avoiding conversational filler unless an explanation is explicitly requested.
Agent Pulse Setup
To deploy Agent Pulse within your environment, ensure the utility scripts are available in your path and configure your agent to recognize the standard triggers.
# Install the pulse evaluation and rendering scripts
cp ./scripts/pulse_eval.py /usr/local/bin/
cp ./scripts/render_pulse.py /usr/local/bin/
# Test the pulse locally
python3 scripts/pulse_eval.py --signal-json '{"active_tasks": 0, "blocked": false}'
Agent Pulse Data Schema & Taxonomy
The skill follows a strict fixed output contract to ensure machine readability and user clarity. The data is organized into the following taxonomy:
| Field | Allowed Values | Description |
|---|---|---|
| status | idle, light, busy, blocked, unknown | The current workload state of the agent. |
| interruptibility | high, medium, low | How easily the agent can pivot to a new request. |
| acceptNewTask | yes, caution, no | A direct indicator for task assignment logic. |
| reason | string (max 12 words) | A brief justification for the current status. |
Internal logic relies on the openclaw-signals.md reference to map runtime signals to these specific outputs.
Agent Pulse Advanced Features
- Script-based Evaluation: Uses dedicated Python scripts (pulse_eval.py) to ensure deterministic results over fluctuating model judgment.
- Low-Cost Signal Processing: Prioritizes visible runtime signals like pending background exec runs over expensive introspection.
- OpenClaw Integration: Features a specific signal-mapping recipe designed for OpenClaw-style runtimes to maximize compatibility.
- Guardrail Protection: Includes logic to prevent the pulse query itself from being counted as workload evidence, preserving data integrity.
SKILL.md
Loading
Related Openclaw Skills
Servo
A local, zero-dependency reference documentation tool for developers providing instant access to technical patterns and guides.
Paper Deep Reader
A high-rigor research assistant designed to deconstruct technical papers into durable, evidence-based markdown notes and implementation memos.
Piston Reference Skill
A local, high-speed reference tool for developers to access Piston concepts, implementation patterns, and best practices directly within their AI agent workflow.
Code → PRD
A powerful reverse-engineering tool that transforms frontend, backend, or fullstack codebases into comprehensive, business-readable Product Requirements Documents.
Greg Eisenberg Startup Ideas & AI Strategy
A high-leverage skill for generating viral business ideas and AI-powered growth strategies inspired by Greg Eisenberg.
Brand Monitoring
A comprehensive tracking tool that monitors brand mentions across social media and news platforms with built-in sentiment analysis and crisis detection.
Featured*
