Ucp Agentic Commerce

PassAudited by ClawScan on May 13, 2026.

Overview

This skill appears to be a legitimate UCP commerce development guide; it does not install hidden code or require credentials, but it can guide agents to browse docs and implement payment/inter-agent workflows.

This looks safe to install as a UCP implementation reference skill. Before using it on real commerce systems, keep web searches and command execution user-approved, use test credentials, and carefully review any generated code that handles payments, OAuth, buyer data, webhooks, or inter-agent communication.

Findings (4)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

Using the skill may cause the agent to browse external UCP, GitHub, Google, or Shopify documentation before writing code.

Why it was flagged

The skill directs the agent to use web/search tools before implementation. This is relevant user-visible behavior, but it is disclosed and aligned with the stated need to track an evolving protocol.

Skill content
Always web-search for the latest specification version and SDK releases before coding.
Recommendation

Allow web access only when appropriate for the project, avoid sharing proprietary details in searches, and pin the UCP spec version used in generated code.

What this means

If you ask the skill to run conformance tests, the agent may propose local commands and external test dependencies.

Why it was flagged

The conformance-testing guidance includes user-directed cloning, dependency installation, and running a test suite. This is central to validation work and not automatic in the artifacts.

Skill content
Clone the conformance repo ... Install dependencies with `uv sync` ... Run tests pointing at your server
Recommendation

Review and approve any clone/install/run commands, run tests in a sandbox or staging environment, and inspect the upstream test repository before executing it.

What this means

Generated implementations may involve payment credentials, OAuth credentials, or buyer identity data if the user builds those features.

Why it was flagged

The skill covers implementations that handle payment tokens and credential-provider flows. This is expected for agentic commerce and the text includes appropriate handling guidance.

Skill content
Credential Provider (CP): Manages payment instruments and user data. Issues payment tokens ... Credentials flow from platform to business ONLY.
Recommendation

Use scoped test credentials during development, store secrets in environment variables or vaults, and require explicit human approval before any real payment flow.

What this means

If implemented, checkout data may move between agents, MCP servers, or embedded checkout frames.

Why it was flagged

The skill includes inter-agent commerce patterns where checkout and payment-related data may pass between agents. This is within the stated scope and other artifacts describe headers, signatures, and origin checks.

Skill content
A2A (Agent-to-Agent) is a protocol for autonomous inter-agent communication. UCP's A2A binding lets a Platform agent talk to a Business agent
Recommendation

Validate agent identities, origins, signatures, and authorization boundaries; test inter-agent and embedded flows with non-production data first.