# Build A2A Protocol Agents

Teams need a practical path to connect agents across frameworks. This skill guides A2A servers, clients, Agent Cards, streaming, and orchestration.

## Install

```bash
npx skillstore add internet-court/a2a-protocol
```

## Metadata

- Status: approved
- Slug: internet-court-a2a-protocol
- Version: 1.0.0
- Author version: 1.0.0
- Skillstore revision: r1
- Version status: valid
- Tree hash: 5f78e96ad6f27f9ff25ed9d3a267c1e83c8a9a297b22785d9f0b5b69338f36ac
- Author: internet-court
- GitHub username: internet-court
- License: Apache-2.0
- Repository: https://github.com/internet-court/internet-court-skill/tree/main/vendored/terminalskills/a2a-protocol
- Ref: 3f6e026a3363e0954ede7bef0cfe88d4475de137
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network, filesystem
- Quality score: 71
- Public page: https://skillstore.pages.dev/skills/internet-court-a2a-protocol
- Manifest: https://skillstore.pages.dev/api/skills/internet-court-a2a-protocol/manifest

## Capabilities

- Outline A2A core concepts, including Agent Cards, tasks, messages, and artifacts.
- Provide Python examples for A2A server setup, task execution, streaming, and cancellation.
- Provide Python client examples for Agent Card discovery and message sending.
- Provide Node.js examples for A2A server and client setup.
- Show sequential and parallel orchestration patterns for multiple A2A agents.
- Compare A2A and MCP so users choose the right integration pattern.

## Use Cases

- Create an Agent Service: Build a focused A2A server with an Agent Card, task handler, streaming updates, and cancellation support.
- Connect Existing Agents: Add A2A clients that discover remote Agent Cards and send tasks to agents built on different frameworks.
- Plan Multi-Agent Workflows: Design routing, sequential delegation, and parallel fan-out across specialized agents while preserving task context.

## Prompt Templates

### Create a Basic Server

```
Build a simple A2A server for a research assistant. Include the Agent Card, one skill, task handling, and cancellation.
```

### Add a Client

```
Create an A2A client that discovers a remote Agent Card, sends a user task, and reports the returned task status.
```

### Support Streaming Tasks

```
Extend the A2A implementation with streaming status updates, input-required handling, cancellation, and clear error states.
```

### Design Agent Orchestration

```
Design an A2A orchestrator that routes tasks across research, writing, and review agents with retries and human handoff.
```

## Limitations

- Examples are implementation templates, not complete production services.
- Users must add real business logic, authentication, authorization, and monitoring.
- The skill does not verify live endpoint behavior or SDK compatibility.
- A2A is evolving, so users should confirm current protocol and SDK versions.

## Best Practices

- Keep each A2A agent focused on one capability domain.
- Declare authentication, supported modalities, and skills clearly in the Agent Card.
- Use streaming, cancellation, and input-required states for long-running tasks.

## Anti Patterns

- Do not expose an A2A server publicly without authentication and authorization.
- Do not create broad agents with vague skills that other agents cannot evaluate.
- Do not use A2A when a simple tool call through MCP is enough.

## Security Audit

- Audited at: 2026-07-21T02:24:39.818\+00:00
- Summary: Most static detections are documentation-format false positives: Markdown fences, example.com URLs, and HTTP discovery routes. One confirmed deployment concern remains: the Python example binds an unauthenticated A2A service to all network interfaces. No prompt-injection text, credential exfiltration, or malicious intent was found.

## Stats

- Views: 0
- Downloads: 4
- Favorites: 1
- Popularity score: 0