AI 交付大脑:给 Agent 用的全生态交付编排 MCP 工具,把一个产品合法打包进 Linux/Windows/苹果/鸿蒙各生态
DeliverKit defines 11 tools with partial schema coverage and inconsistent descriptions. Tool names follow verb_noun convention (inspect_project, generate_packaging_plan, pack_deb, etc.), which is good. However, descriptions vary widely in quality: some are specific (e.g., pack_deb: 'build deb package in isolated Ubuntu container'), others are vague or in Chinese without English translation. Input schemas are present but lack complete type definitions for all parameters, many parameters have descriptions but no explicit 'type' field visible in the schema. Output schemas are not documented. Error handling is minimal; no recovery guidance or actionable error messages visible. The server lacks tool annotations (readOnlyHint, destructiveHint) despite having clear read-only vs. write operations. No pagination support visible for list-like operations. Overall, this is a functional but incomplete implementation typical of community servers.
按 Forge.md 生成 GitHub Actions 多平台交付工作流
生成 Forge.md 交付契约(目标生态 / 产物 / 决策 / 风险)
汇总各平台 JSON 结果并生成 ReleaseManifest.json
查询特定生态的交付知识库(目标、约束、工具链、验证方法)
识别项目语言、入口与已有打包配置,给出跨生态交付目标建议
按 Forge.md 在隔离 Ubuntu x86_64 容器中构建 AppImage,并执行 extract-and-run 验证
按 Forge.md 在隔离 Ubuntu 容器中构建 deb 包,并执行干净容器安装/运行验证
在 DevEco runner 构建并使用 hdc 安装验证 HarmonyOS HAP/APP
在 macOS runner 构建、签名、公证并验证 DMG/PKG
Descriptions contain Chinese text without English translation (e.g., '识别项目语言、入口与已有打包配置'). LLMs trained primarily on English will not understand tool purpose. All descriptions must be in English.
Output schemas are not documented. LLMs cannot plan downstream tool calls or extract required fields (e.g., what does pack_deb return? Does it include artifact_path, checksum, size?). Every tool must document its return type.
Tool annotations (readOnlyHint, destructiveHint, idempotentHint) are missing. inspect_project and get_ecosystem_knowledge are read-only; pack_* and generate_* are destructive/write operations. Annotations help LLMs reason about side effects and retry safety.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2026-07-28+ | v2 |
按 Forge.md 在隔离 Rocky Linux 容器中构建 rpm 包,并执行干净容器安装/运行验证
在 Windows runner 使用 WiX 构建、签名并验证 MSI
Parameter descriptions lack format constraints and validation rules. E.g., 'goals' parameter accepts 'comma-separated' values but no enum or pattern is defined. 'env' parameter lists examples (ubuntu-22.04, windows, macos, harmonyos) but no formal constraint. LLMs will hallucinate invalid values.
No error handling or recovery guidance visible. If pack_deb fails (missing dependencies, invalid Forge.md, container unavailable), what should the LLM do? Retry? Call a different tool? Provide actionable error messages and recovery paths.