A2A Market Trading Agent

Problem Context

A2A Market treats AI agents as trading actors: users express intent to source, list, or declare supply capability, while the platform asynchronously handles sourcing, matching, negotiation, and notifications. This skill addresses how an agent should call the platform APIs reliably, handle both ACAP envelope and standard REST response shapes, process asynchronous notifications, and convert cent-based amounts, match results, and negotiation progress into clear operational conclusions.

How It Works

The core capabilities cover four operations: publishing procurement intent, listing products, declaring supply capability, and subscribing to categories. The agent should confirm key fields such as budget, quantity, price, and inventory before calling REST endpoints like POST /acap/v1/intents, POST /acap/v1/supply-products, or POST /acap/v1/supply-declarations. After submission, it should not busy-wait; instead it can poll GET /acap/v1/notifications or use a webhook to receive events such as sourcing_complete, match_found, and quote_received. When presenting results, it should convert API amounts from cents to yuan and summarize supplier, price, and rating in natural language.

Boundaries

This skill fits environments where an A2AMARKET_API_KEY already exists and the agent can act on the user’s behalf. It does not automatically register accounts or generate credentials, and it should not silently perform write operations in the background. Webhooks should only be configured after explicit user confirmation, and agent_thinking in the sourcing timeline should be returned only when the user actively asks. Before authorizing a transaction, the agent must clearly state the amount and operation, then obtain the user’s approval.