Agpair is a command-line control surface that allows AI agents to delegate execution tasks to Antigravity while maintaining strict oversight and health monitoring.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agpair
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 agpair 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 Agpair?
Agpair serves as the essential bridge between a primary AI controller and the Antigravity execution engine. Within the ecosystem of Openclaw Skills, it functions as a specialized interface for dispatching coding work, monitoring daemon health, and managing task lifecycles without overstepping into semantic decision-making. It ensures that the primary agent remains the reviewer and controller, while Antigravity handles the heavy lifting of code execution.
By leveraging this skill, developers can implement a clear separation of concerns in their AI workflows. The tool provides robust mechanisms for preflight checks, ensuring that repositories are healthy and bridge sessions are ready before any code is modified. This disciplined approach prevents common pitfalls in automated coding environments, such as conflicting reader states or orphaned processes.
Agpair Use Cases
- Delegating complex coding tasks from a primary AI agent to Antigravity for execution.
- Performing preflight diagnostic checks on repositories to ensure environment health.
- Monitoring long-running background tasks through real-time status updates and log tailing.
- Executing semantic follow-up actions like continuing, approving, rejecting, or retrying delegated work based on evidence review.
How Agpair Works
- Preflight Diagnosis: The system runs health checks on the repository and daemon status to ensure the environment is ready for task delegation.
- Task Dispatch: Tasks are dispatched to Antigravity using the CLI, which returns a unique Task ID for tracking.
- Active Monitoring: The controller monitors the task using status and log commands, maintaining a blocking wait if necessary to ensure process completion.
- Evidence Review: Upon task completion or a terminal phase, the controller reviews the EVIDENCE_PACK to verify the quality of the work.
- Semantic Finalization: The controller issues a final command (approve, reject, or continue) to close the loop on the delegated task.
Agpair Setup
To integrate this component of Openclaw Skills into your environment, ensure the agpair CLI is installed and accessible in your path. Configure your primary AI agent to recognize the following triggers.
# Check if the environment is ready
agpair doctor --repo-path /absolute/path/to/repo
# Ensure the background daemon is running
agpair daemon status
# Dispatch a new task
agpair task start "Your task description here"
Agpair Data Schema & Taxonomy
Agpair organizes its operation around Task IDs and specific state flags to maintain technical truth during execution.
| Attribute | Description |
|---|---|
| TASK_ID | Unique identifier for every delegated unit of work. |
| desktop_reader_conflict | Boolean flag indicating if a local environment conflict exists. |
| repo_bridge_session_ready | Status of the connection between the controller and the executor. |
| waiter_state | Current state of the polling process (e.g., waiting, exited). |
| EVIDENCE_PACK | The collection of logs and diffs produced by the execution phase. |
Agpair Advanced Features
- Multi-agent coordination where one agent acts as a supervisor and another as the executor via Antigravity.
- Automated blocking wait discipline that prevents premature task abandonment.
- Semantic follow-up controls (continue, approve, reject, retry) for granular lifecycle management.
- Detailed log-level inspection with configurable limits for deep debugging within the CLI.
SKILL.md
Loading
Related Openclaw Skills
Package Manager Updates
A comprehensive automation tool to check, summarize, and update outdated packages across multiple package managers including npm, pip, and brew.
Refactoring (Deep Workflow)
A comprehensive deep refactoring workflow designed to restructure code safely through characterization tests, incremental steps, and behavior preservation.
Claw Search
A zero-config, free search engine API designed to give AI agents real-time web, image, and news retrieval capabilities.
Whisper-Piper Voice Pipeline
A fully local, high-performance speech-to-text and text-to-speech HTTP server designed for privacy-first voice assistants.
Quantum Technology Career Roadmap
A personalized guidance platform that generates structured learning roadmaps for professionals and students entering the quantum technology sector.
Visual Note Prompt Generator
A professional AI prompt engineer designed to transform brief text into standardized, visual knowledge card prompts.
Featured*
