stdio(推荐)
weapp-vite mcpweapp-vite ai console
以三条主线组织:MCP 连接能力、Skill 统一工程实践、LLMS 喂给上下文。
MCP
先接入服务,再让 AI 调 tools,避免纯口头建议。
weapp-vite mcpweapp-vite mcp --transport streamable-http --host 127.0.0.1 --port 3088 --endpoint /mcpSkill
区分公开 skills 与项目本地附加 skills,让团队在稳定边界下协作与交付。
npx skills add sonofmagic/skillspnpm skills:link$weapp-vite-best-practices
$docs-and-website-sync
$github-issue-fix-workflow
$release-and-changeset-best-practices
$weapp-devtools-e2e-best-practices
$weapp-vite-wevu-performance-best-practices
$weapp-vite-vue-sfc-best-practices
$wevu-best-practices
$native-to-weapp-vite-wevu-migration
$weapp-ide-cli-best-practices$playwright-cli`skills/*` 面向公开分发;`.claude/skills/*` 面向这个仓库的本地附加能力,并通过 `pnpm skills:link` 一起同步。
$weapp-vite-best-practices$docs-and-website-sync$github-issue-fix-workflow$release-and-changeset-best-practices$weapp-devtools-e2e-best-practices$weapp-vite-wevu-performance-best-practices$weapp-vite-vue-sfc-best-practices$wevu-best-practices$native-to-weapp-vite-wevu-migration$weapp-ide-cli-best-practices$playwright-cliLLMS
按“索引 -> 全文 -> 结构化索引”逐层加载,降低噪声。
1) /llms.txt
2) /llms-full.txt
3) /llms-index.json
4) 结合 MCP tools 继续定位与验证