Preflight

io.github.evanfollis由 GitHub 账号 evanfollis 发布v0.1.1更新于 Oct 24, 2026

Check if your MCP server is ready to publish on the MCP Registry, Smithery, or npm.

发布方自己的服务：Preflight

已验证Streamable HTTP可网页运行开发者工具

概览

AI 生成的概览

Preflight 是一个远程 MCP 服务器，用于检查某个 MCP 服务器是否已具备发布到 MCP Registry、Smithery 或 npm 的条件。

可以做什么
校验 MCP 服务器能否发布到 MCP Registry、Smithery、npm 等真实目录，并给出有证据支撑、关联目录规则来源的检查结果。
适用场景
在把 MCP 服务器提交到 MCP Registry、Smithery 或 npm 之前使用，用来发现发布准备问题，并了解包不符合哪些目录规则。
运行要求
需要远程 Streamable HTTP 端点；未声明任何软件包、环境变量或请求头，也未声明身份验证。
安装前请注意
该工具会返回原始产物内容，分享结果前请先检查其中内容；输入中未显示凭据、付费或写入操作。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Preflight，将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Web executable，通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

交给智能体安装

把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体，或 SourceWeft 对话中。智能体会读取该服务器的安装指南，先向你展示发布者、端点和登录方式，经你同意后再把它添加到你的客户端。密钥不会写进任何命令。

阅读 https://sourceweft.com/mcp/io.github.evanfollis%2Fpreflight/install.md，并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前，先告诉我它的发布者、端点和登录方式，等我确认。不要把 API 密钥写进命令或我会提交的文件；问我密钥应从哪里获取。

查看智能体所依据的安装指南

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。

{
  "mcpServers": {
    "preflight": {
      "type": "http",
      "url": "https://skillfoundry.synaplex.ai/products/preflight/mcp/"
    }
  }
}

README

skillfoundry-products

skillfoundry-products is the implementation repository for the probes, mechanisms, and offerings that Skillfoundry builds and operates.

Lifecycle: active. This monorepo is the canonical source for its maintained products, including Preflight.

Verified today: TypeScript workspace checks and both Python suites run through the root gate. Registry publication is separate and occurs only from pinned, checksum-verified OIDC workflows.

bash
make setupmake check

See docs/architecture.md for ownership boundaries, artifact roles, and dated migration exceptions.

This repo has two different lanes on purpose:

  • mechanisms/: internal factory tools that help Skillfoundry identify, design, and improve opportunities. These are not automatically public products.
  • products/: external, user-facing software artifacts that have been selected for public deployment, distribution, and monetization.
  • templates/: reusable implementation starters for future external product lanes.

Public product routing should stay platform-centric. The preferred external shape is:

  • https://skillfoundry.synaplex.ai/products/<product-slug>/...

not one public hostname per product.

The distinction matters. Internal mechanisms help the agent system think and decide better. External products are what we eventually sell or distribute through channels such as AgenticMarket.

Default External Stack

The default lane for new external MCP offerings or probes is now:

  • TypeScript-first
  • remote MCP over Streamable HTTP
  • Cloudflare Workers for stateless offerings
  • Skillfoundry-owned public routes for the canonical runtime surface

That keeps the runtime independent from any single marketplace while preserving a simple path to distribution through registries and marketplaces later.

Current Layout

  • mechanisms/bottleneck-radar/: internal MCP mechanism for turning noisy source signals into ranked bottlenecks and builder-ready opportunity briefs.
  • products/launchpad-lint/: first external MCP product candidate for auditing launch readiness and drafting marketplace launch packages.
  • templates/remote-skill-ts/: default starter for new remote TypeScript MCP probes or offerings.

Working Rule

Do not treat a mechanism as an offering by default.

A mechanism graduates into an offering only when Skillfoundry has made an explicit decision that:

  • the user-facing utility is clear,
  • the deployment surface is chosen,
  • the pricing or monetization path is chosen,
  • and the artifact deserves its own external lifecycle.

Factory Documents

  • docs/CASHFLOW_ENGINE.md: repo-level operating loop for commercial probes and offerings
  • docs/BUILD_NEXT_PORTFOLIO.md: current ranking of the next revenue lanes
  • docs/LEARNING_LAB.md: standard for turning launches and failures into reusable learning
  • docs/HYPOTHESIS_TEMPLATE.md: required hypothesis format for major changes
  • docs/SKILL_FACTORY_SCORECARD.md: build-next scoring heuristic
  • docs/TELEMETRY_MODEL.md: minimum telemetry and evidence envelope for compounding learning
  • docs/SKILL_PROPOSAL_TEMPLATE.md: required pre-build probe template
  • docs/PORTFOLIO_INTAKE.md: intake workflow for new product and suite ideas before they graduate into active builds
  • docs/POST_LAUNCH_REVIEW_TEMPLATE.md: required after-launch learning template
  • docs/ADVERSARIAL_REVIEW_PRIMITIVES.md: multi-lens skeptical review primitive for sharpening products and extracting reusable heuristics

TypeScript Workspace Setup

The repo now includes a pnpm workspace for future TypeScript skills. If pnpm is not installed globally, use corepack:

bash
corepack pnpm install

来源：README.md，提交 b35b0d5

工具

1

check_publish_readiness

Validate whether an MCP server is publishable on real directories (MCP Registry, Smithery, npm). Provide raw artifact contents. Returns evidence-backed findings with source-linked directory rules.

输入参数readmemanifestpackage_jsonsmithery_yamlpyproject_tomltarget_directories

版本历史

1
  1. v0.1.1最新Sep 16, 2026

评分

已满足适用检查项的 90%

该评分衡量此收录条目的完整程度，依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价，也不表示服务器安全或当前可用：可连通性是带日期的检查结果，评分也不包含安装量或用户评价。

  • 近 30 天内可连通必需25 分
  • 已声明认证方式必需10 分
  • 已列出工具必需15 分
  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分

    请在仓库或注册中心条目中声明许可证。

其他 Preflight 集成

Preflight

social.preflight官方发布者

托管式社交媒体排程服务器，让助手起草、按规则校验、排程并向 12 个平台发布帖子。

Streamable HTTP可网页运行鉴权
效率与工作流需要凭据Oct 27, 2026

SQLite rebuild copy-map preflight

由 GitHub 账号 showflexinc 发布

远程预检服务，在重建执行前检查 SQLite 表重建复制映射的结构。

Streamable HTTP可网页运行鉴权
数据库需要凭据Nov 4, 2026

Rust MCP Preflight

由 GitHub 账号 Wolfe-Jam 发布

在发布前检查 MCP Registry 条目：README 中的 mcp-name 行、server.json 的发布规则以及服务器名称。

STDIO仅桌面
开发者工具尚未验证Nov 7, 2026

Multipart row preflight

由 GitHub 账号 showflexinc 发布

CSV row-structure inspection (inspect_parts) and manifest comparison (compare_manifest).

Streamable HTTP可网页运行鉴权
数据与分析需要凭据Oct 31, 2026

Tool-dialogue preflight

由 GitHub 账号 showflexinc 发布

针对JSONL工具对话训练数据的结构化预检。AI 翻译

Streamable HTTP可网页运行鉴权
需要凭据Oct 31, 2026