WeChat Mini Program AI Collaboration Rules Generator (AGENTS.md)

Problem This Skill Addresses

When AI assists with WeChat Mini Program development, the failure mode is often not missing syntax, but uncontrolled behavior: refactoring unrelated modules, embedding secrets in code, running irreversible actions without confirmation, or missing reference updates after renames. In mini program projects, the risk grows with subpackages, Canvas, locked templates, and multiple feature boards.

How It Works

The skill turns collaboration rules into a generated and inspectable project file:

• Renders an AGENTS.md at the project root for AI to read at the start of each session.
• Encodes baseline rules for safety, stability, communication, minimal changes, and reference synchronization, with high-priority restrictions on payments, deletion, .env, and Git commits.
• Adds stronger constraints for subpackages and Canvas work, such as naming isolation, core template locking, Canvas sizing, and draw-order rules.
• Supports --check to scan for hard-coded secrets or .env files in version control.
• Uses --self-test to re-read the generated rules and verify required principles are present.

Boundaries and Notes

This skill provides a general collaboration baseline, not project-specific MVP scope, API inventories, or business documentation indexes. The default tech stack assumptions cover WeChat Mini Program, FastAPI, and a database; change the stack-specific section when migrating. Human review and delivery acceptance remain necessary.