loop-engineer
Engineers multi-skill systems by auditing and building linked skill packages.
Install
mkdir -p .claude/skills/loop-engineer && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/18818" && unzip -o skill.zip -d .claude/skills/loop-engineer && rm skill.zipInstalls to .claude/skills/loop-engineer
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Loop 系统工程师 — 从用户需求出发，设计并开发完整的多 skill 联动 package。扫描现有 skill 资产，识别可复用与缺失项，逐一开发后组包，编写主调度 skill 平滑层。也可用于对已有 package 进行联动完整性审计（格式合规+主调度逻辑+入口文档同步+命名一致性）。当用户说"我需要一个 XX 系统/loop/agent 包"、"帮我设计一个多技能联动方案"、"检查 package 联动完整性"时触发。不适用于单个 skill 开发。Key capabilities
- →Design multi-skill linked packages from user requirements
- →Scan existing skill assets for reusability and gaps
- →Develop missing skills using `skill-governor`
- →Assemble skills into a package structure
- →Write a main dispatching skill (smoothing layer)
- →Audit existing packages for linkage integrity
How it works
The skill interprets user requirements, inventories existing skills, develops missing components, and then integrates them into a package with a main dispatching skill. It also audits existing packages for consistency.
Inputs & outputs
When to use loop-engineer
- →Design agent system
- →Audit package integration
- →Build multi-skill loop
About this skill
Loop Engineer — 需求驱动的系统化 Skill Package 工程师
Overview
Loop Engineer 是 skill-governor 的上层调度者。从用户需求出发，设计完整的多 skill 联动体系：盘点现有资产、识别缺口、逐一补齐、组包联调、编写平滑层，最终交付完整 package。
两种工作模式：
- 模式 A: 新建 Package — 完整执行 Phase 0 → Phase 5 全流程
- 模式 B: 审计已有 Package — 执行 Phase 1 资产盘点 + Phase 5 验证 checklist，输出诊断报告
Workflow
Phase 0: 需求理解与 OODA 映射
- 接收用户需求，提炼能力模块清单
- 映射 Observe/Orient/Decide/Act 四槽位 + 修正面
- 获得用户确认
Phase 1: 资产盘点与 Gap 分析
- 扫描现有 skill 列表
- 生成 Gap 报告（可复用 / 需适配 / 需新建）
- 【必须】识别主调度 Skill — 每个 package 必须有且仅有一个主调度节点
- 【必须】Package 定位输出 — 目的 + 主调度名 + 子 skill 职责
- Gate: 用户确认
Phase 2: 缺失 Skill 开发
- 对每个缺失 skill 走 skill-governor 全流程（三镜像同步）
- Gate: 缺失列表清零
Phase 3: 组包
- 创建 package 三平台目录
- 复制所有相关 skill
- 路径验证
Phase 4: 平滑层与主调度
- 确定主调度 skill（可复用已有 / 新建 orchestrator / 加前置分诊）
- 编写主调度 SKILL.md（能力索引 + 路由逻辑表 + 数据流 + 回流机制）
- 上下文传递与错误回退
Phase 5: 验证与交付
5.1 三平台格式合规检查 — Codex/Claude/Gemini 格式 + description 语义一致
5.2 主调度逻辑检查 — 有主调度 / 路由表完整 / 无歧义 / 有回流机制
5.3 入口文档同步检查 — CLAUDE.md / AGENTS.md / GEMINI.md 列出所有 skill 且命名一致
5.4 命名一致性检查 — command 名可追溯到 skill 名，无混淆或重复
5.5 联动路径检查 — 引用路径存在 / scripts 有效 / 接口对齐 / 无孤立文件
5.6 README 生成/更新 — 架构图 + 清单 + 主调度说明 + 部署 + 示例
Gate: 用户确认 → commit
Output Contract
完成时必须产出：Gap Analysis Report（含主调度识别）、完整 Package 目录（三平台）、主调度 Skill（含路由表+回流机制）、平滑层、入口文档（CLAUDE.md/AGENTS.md/GEMINI.md）、Package README。
Resources
| 文件 | 用途 | 何时加载 |
|---|---|---|
| skill-governor | 单 skill 开发规范 | Phase 2 |
| 项目 README | 现有 skill 清单速查 | Phase 1 |
关联 Skill（网络调度协议）
| 关系 | Skill | 场景 |
|---|---|---|
| 被调用 | ai-spec | 需要设计多 skill 联动系统时 |
When not to use it
- →When developing a single skill
- →When the user does not need a multi-skill package or integration audit
Limitations
- →Not applicable for single skill development
- →Each package must have one main dispatching skill
- →Requires user confirmation at various gates
How it compares
This skill provides a structured methodology for designing and auditing multi-skill packages, ensuring complete integration and consistency, unlike ad-hoc skill development.
Compared to similar skills
loop-engineer side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| loop-engineer (this skill) | 0 | 26d | No flags | Advanced |
| meta-automation-architect | 7 | 8mo | Review | Advanced |
| hive-mind-advanced | 6 | 4mo | Review | Advanced |
| babysitter:assimilate | 0 | 4mo | No flags | Advanced |
Try saying
Example prompts that trigger this skill in your AI assistant.
More by PancrePal-xiaoyibao
View all by PancrePal-xiaoyibao →You might also like
meta-automation-architect
comzine
Use when user wants to set up comprehensive automation for their project. Generates custom subagents, skills, commands, and hooks tailored to project needs. Creates a multi-agent system with robust communication protocol.
hive-mind-advanced
ruvnet
Advanced Hive Mind collective intelligence system for queen-led multi-agent coordination with consensus mechanisms and persistent memory
babysitter:assimilate
MaTriXy
Assimilate an external methodology, harness, or specification into babysitter process definitions.
autonomous-agents
davila7
Autonomous agents are AI systems that can independently decompose goals, plan actions, execute tools, and self-correct without constant human guidance. The challenge isn't making them capable - it's making them reliable. Every extra decision multiplies failure probability. This skill covers agent loops (ReAct, Plan-Execute), goal decomposition, reflection patterns, and production reliability. Key insight: compounding error rates kill autonomous agents. A 95% success rate per step drops to 60% b
chief-architect
ananddtyagi
PERSONAL APP ARCHITECT - Strategic development orchestrator for personal productivity applications. Analyzes project context, makes architectural decisions for single-developer projects, delegates to specialized skills, and ensures alignment between user experience goals and technical implementation. Optimized for personal apps targeting 10-100 users.
agent-planner
ruvnet
Agent skill for planner - invoke with $agent-planner