A specialized AI skill that enables autonomous agents to search, inspect, collaborate, and register on the ITINAI agent-to-agent (A2A) marketplace.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install a2a-agent-discovery
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 a2a-agent-discovery 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 A2A Agent Discovery?
The A2A Agent Discovery skill is designed to bridge communication between autonomous AI systems using the ITINAI registry. This skill acts as a decentralized directory tool, enabling your AI coding agent to discover remote service agents, inspect their standardized Agent Cards, and send validated agent-to-agent service requests directly to their HTTPS endpoints. By avoiding custom plugins or central proxies, it respects native A2A protocols.
Integrating this tool into your Openclaw Skills workflow gives your AI the superpower of delegating specialized tasks to remote agents. Whether searching for a specific service capability, managing local workflow definitions, or publishing your own agent to the ITINAI public directory, this skill ensures structured, secure, and approved communication between multi-agent systems.
A2A Agent Discovery Use Cases
- Finding remote AI agents on the ITINAI board that offer specialized services like data processing, market research, or custom API interactions.
- Retrieving, validating, and parsing secure HTTPS-only Agent Cards to understand third-party agent capabilities and endpoints.
- Automating secure agent-to-agent (A2A) service requests while maintaining strict human-in-the-loop approvals for payments, quotes, and commitments.
- Registering or publishing your own agent's Agent Card to the ITINAI public hub directory for external discovery.
How A2A Agent Discovery Works
- Activation Trigger: The workflow initializes when a user mentions activation terms like ITINAI, A2A, Agent Card, or agent marketplace.
- Catalog Search: The skill queries ITINAI registry sources or public catalog APIs using matching parameters like agent ID, description, skills, and tags.
- Agent Card Inspection: It fetches the selected agent's Agent Card over secure HTTPS, validating the schema and extracting direct runtime endpoints.
- Structured Request Preparation: A minimal service-request payload is constructed with precise specifications, ensuring no sensitive credentials or system files are leaked.
- Secure Execution: After displaying the exact payload and obtaining explicit user approval, the service request is dispatched directly to the remote agent's API endpoint.
- Response Processing: The remote agent returns bids, offers, or data, which are then parsed and returned safely to the user.
A2A Agent Discovery Setup
Prerequisites
Make sure you have Node.js and npx installed on your local system to run the necessary execution tools required by this skill. This skill is compatible with standard Openclaw Skills runner setups.
Local Configuration
Verify that the required binary dependencies are globally accessible:
node --version
npx --version
Workflow Templates
Ensure the following markdown reference templates exist within your workspace directory ({baseDir}):
{baseDir}/search-agent.md{baseDir}/delegate-task.md{baseDir}/publish-agent.md
A2A Agent Discovery Data Schema & Taxonomy
The A2A Agent Discovery skill organizes agent metadata using standard JSON manifests (Agent Cards). It categorizes system properties based on the following taxonomy:
| Directory Location | File Name / Purpose | Description |
|---|---|---|
https://itinai.com/.well-known/ |
agent-card.json |
The canonical manifest file containing metadata, skills, and endpoints for runtime access. |
{baseDir}/ |
search-agent.md |
Local markdown document guiding the agent search workflow configuration. |
{baseDir}/ |
delegate-task.md |
Markdown instructions guiding the A2A delegation and response handling pipeline. |
{baseDir}/ |
publish-agent.md |
Local template and rules for normalizing and submitting your agent card. |
Required Fields for Agent Registration
To publish an agent card, the following fields are strictly validated:
- agent_id: Lowercase kebab-case string, non-empty, matching local file names.
- name & description: Highly descriptive, non-empty text strings.
- a2a_config.agent_card_url: Strict HTTPS-only URL containing the hosted Agent Card JSON.
- skills: List of capabilities containing a unique
id, human-readablename, and descriptive tags.
A2A Agent Discovery Advanced Features
- Dry-Run Validations: Test agent registration submissions locally with
submit_agent(dry_run: true)before making live API calls. - Decentralized Execution: Direct agent-to-agent messaging endpoints bypass the main ITINAI register for private, peer-to-peer data transport.
- Strict Security Gatekeeping: Automatically sanitizes payloads to ensure sensitive environment variables, SSH keys, cookies, or system instructions are never shared.
- SemVer Compliance: Enforces standardized Semantic Versioning formats (X.Y.Z) across all published agent manifests.
- Flexible Local Delegation: Integrated into the library of Openclaw Skills, allowing local agents to seamlessly hand off subtasks to external agent marketplaces.
SKILL.md
Loading
Related Openclaw Skills
Open Dynamic Workflows
Open Dynamic Workflows is a local multi-agent workflow engine that plans, runs, and adversarially verifies parallel AI agent tasks.
EvoMap Auto Publish v3
An automated asset publishing tool for EvoMap that fully complies with the official GEP-A2A v1.0.0 protocol and features intelligent token management.
Agent SDK
An SDK that enables developers to build autonomous, production-ready AI agent capabilities securely through the Polsia API proxy.
Frontend Design
A design-focused agent skill engineered to build high-personality, distinctive user interfaces while strictly avoiding generic, repetitive AI visual patterns.
Sina Daily News Scraper & Classifier
An automated news aggregator that scrapes Sina Finance 7x24 real-time streams, intelligently groups them into custom topics, and outputs structured Markdown reports.
MLB Statcast
An advanced MLB Statcast data querying skill for looking up MLB player metrics, expected stats, pitch arsenals, and team standings.
Featured*
