wechat-devtools-control

name	wechat-devtools-control
description	Use this skill for end-to-end WeChat DevTools workflow control in this repo, including bootstrap, environment doctor checks, natural-language project creation, generated project preview/guard/upload dry-run, validation gates (fast/full), and guarded deploy orchestration when no narrower WeChat skill applies.

WeChat DevTools Control

Use this fallback umbrella skill for the local WeChat workflow when a narrower WeChat skill is not explicitly selected.

Entry

$RepoRoot = (Get-Location).Path
. (Join-Path $RepoRoot "scripts\wechat.ps1")

Main workflow

Invoke-WechatBootstrap
Invoke-WechatDoctor
Invoke-WechatCreate -Prompt "build a notebook mini program" -Open $true -Preview $true

Generated project operations

Get-GeneratedProjectList
Invoke-GeneratedProjectOpen
Invoke-GeneratedProjectPreview
Invoke-GeneratedProjectDeployGuard
Invoke-GeneratedProjectSetAppId
Invoke-GeneratedProjectUpload -DryRun $true

Validation

powershell -ExecutionPolicy Bypass -File (Join-Path $RepoRoot "scripts\test-wechat-skill.ps1") -GuardCheckOnly
powershell -ExecutionPolicy Bypass -File (Join-Path $RepoRoot "scripts\test-wechat-skill.ps1") -SkipSmoke -Tag fast

Safety defaults

• Keep generated projects preview-first by default.
• Block deploy/upload for touristappid projects.
• Allow guarded upload/deploy only after setting a real appid and passing checks.
• Keep broad write/deploy routes disabled unless explicitly requested.

AutoFix loop policy (required)

When the user asks for "generate and make it correct", this skill must follow a repair loop and must not report success early.

1. Generate/update the target project code.
2. Run detector round (automator preferred; fallback route allowed).
3. If any issue is found, produce and apply a repair change.
4. Re-run detector round.
5. Repeat until:
   • success: checks pass, or
   • blocked: runtime/environment blocker, or
   • failed: max rounds reached.

Hard rule:

• Do not return success while known issues remain.
• Do not wait for extra user feedback between loop rounds unless the state is blocked.
• Runtime blockers (for example invalid appid, automator startup failure) must be surfaced explicitly as blocked, not hidden as pass.

Compile error gate (required)

After project open/startup, compile/runtime parse errors in DevTools must be treated as blocking issues, not as startup success.

Examples:

• WXML parse errors such as unexpected token / Bad value with message
• file-level compile errors like at files://...wxml
• JS runtime load failures that stop page render

Required behavior:

1. Run a post-open detection round before declaring success.
2. If compile error evidence exists, classify as issue (critical), enter repair loop, and re-check.
3. Only return success when compile errors are cleared and detector result is healthy.
4. Collect console logs into artifacts\wechat-devtools\console\latest.log and evaluate them in every repair round.

Prefer narrower skills when appropriate

• wechat-spec-executor
• wechat-release-guard
• wechat-lab-builder