WeChat open‑sources payment Skills

WeChat Pay has published an open-source toolkit for artificial-intelligence coding assistants that walks developers through payment integration instead of leaving them to stitch it together by hand. (github.com, pay.weixin.qq.com) The repository, `wechatpay-skills`, appeared on GitHub in late March and had 103 stars and 9 forks when checked on April 15, 2026. Its README says the package organizes payment product knowledge, code examples and integration rules so an artificial-intelligence agent can answer developers with product-specific guidance. (github.com) WeChat Pay’s own documentation page for the basic-payment Skill was updated on April 8, 2026 and says developers can load the package with `npx skills add --yes`, clone the repository, or copy the Skill folder into an integrated development environment such as Cursor. (pay.weixin.qq.com) A “Skill” here is a rulebook for a coding assistant: it tells the assistant what questions to ask, what files to search, and what not to invent. The basic-payment Skill says the assistant must first confirm whether a developer is using merchant mode or service-provider mode, then retrieve existing examples instead of writing payment code from scratch. (github.com, pay.weixin.qq.com) That matters because WeChat Pay’s standard flows already involve multiple moving parts before money changes hands. In JavaScript application programming interface payment, for example, a merchant backend creates an order, receives a prepayment identifier, calls WeixinJSBridge in the WeChat client, then still has to verify the order through query and callback handling after the user returns. (pay.weixin.qq.com) The same pattern shows up in application payment. WeChat’s application-payment guide says merchants must create the order on the backend, use the WeChat Open Software Development Kit to launch checkout in the app, and then confirm the final order state through the query interface and payment callback rather than trusting the client-side return alone. (pay.weixin.qq.com) The new Skill package is built around that complexity. The repository says each Skill covers five jobs: payment-method selection, code-example lookup, business-knowledge lookup, integration quality checks, and troubleshooting for errors such as bad signatures, missing callbacks or refund failures. (github.com, github.com) WeChat Pay’s documentation says the basic-payment Skill covers Java and Go examples, and its examples span JavaScript application programming interface, app, Hypertext Markup Language 5, Native, Mini Program, payment-code and combined-order scenarios. The same page says the assistant can also use a Request-Id from an application programming interface response header to match known error-handling paths. (pay.weixin.qq.com, github.com) The package also reflects a broader change in how WeChat documents its platform for machine-assisted development. Current merchant-doc pages now include “Copy for LLM” and “Ask AI” options, and several pages explicitly tell large language models to consult `llms.txt`, showing the documentation itself is being formatted for artificial-intelligence retrieval as well as human reading. (pay.weixin.qq.com, pay.weixin.qq.com) For developers, the practical shift is narrower than “artificial intelligence writes your payments stack.” WeChat’s own Skill rules say the assistant should only surface prewritten examples, should not write directly into project files, and should check signatures, callbacks and required interfaces before launch. (github.com, pay.weixin.qq.com) So the release is less a new payment rail than a new wrapper around an old one: WeChat Pay is packaging its existing checkout flows into machine-readable instructions, then telling coding assistants to follow the playbook. (github.com, pay.weixin.qq.com)