Adapting WeChat Mini Programs with Evidence Layers

Where this fits

You ask an agent to check why a WeChat Mini Program works in DevTools but fails on a real iPhone. The agent digs into config files, sees a valid appid, and says "should work" — but that is exactly the trap. This skill forces agents to treat source code, build output, DevTools loading, device behavior, permissions, privacy declarations, backend linkage, and release status as separate evidence layers. It applies to both a new idea and an existing repository, and it keeps the agent in read-only mode until you explicitly authorize writes.

Why agents benefit

• Separates evidence layers: The skill never upgrades "code exists" into "DevTools loaded" or "DevTools visible" into "production works." Agents produce honest, staged conclusions.
• Enforces safety boundaries: By default, it inspects only entry points, build directories, and public config. Real AppID, secrets, cloud env identifiers, and credentials are redacted or reported only as hashed fingerprints.
• Checks official sources: For time-sensitive claims like privacy declarations or review rules, the agent verifies against current WeChat official documentation instead of relying on stale training data.
• Defines a clear workflow: From problem classification to read-only inspection to evidence reporting, the agent follows a repeatable process rather than guessing.
• Works across project states: Handles raw ideas, existing repos, platform error messages, and release-readiness questions with the same evidence discipline.

Practical scenarios

Pre-upload repository audit

Before the first upload, the agent walks a fresh repo and classifies what is verifiable: project root, compile paths, app.json permissions, and privacy declarations. It reports which assumptions need manual confirmation in WeChat DevTools.

Device-versus-DevTools mismatch

A developer sees a feature work locally but fail on a real device. The agent separates build output from device behavior, checks permission scopes and API availability, and flags claims that need a physical device test.

Privacy compliance review

Before submission, the agent checks privacy-related configuration and permission declarations against official requirements, distinguishing “declared in source” from “accepted by review.” It never claims compliance from source alone.

Add it to your agent workflow

Install the skill with:

npx skills add NocodeMrLi/mini-program-engineering-skill-suite --skill wechat-mini-program-platform-skill

Then ask something like: “Inspect this Mini Program project and list which platform evidence layers are confirmed, which are unknown, and whether privacy declarations are ready for review.” The agent will inspect read-only and return a layered evidence report.

Compare before adopting

A skill like this should be judged on more than its README. Check the GitHub repository’s issue activity, star trajectory, and whether the author updates it when WeChat platform rules change. Compare it against general-purpose WeChat automation skills that may lack the evidence-layer discipline or read-only safety guarantees. Also consider whether the skill’s workflow fits your agent’s existing project inspection patterns.

Why it is worth tracking

At 22 stars, this is not a mainstream skill yet, but its design is valuable because it solves a real failure mode: agents overstating platform readiness. The author’s focus on evidence separation and security boundaries gives it a distinct identity. It is worth re-evaluating when the repository shows commits tied to WeChat rule changes or when more agents adopt it. For a maintained index of such skills, see https://www.openagentskill.com/skills/nocodemrli-wechat-mini-program-platform-skill.