skill-pipeline-orchestrator
Skill1 starsChanged 4 months agoMITView on GitHub
Use when the user says "一键发布 skill", "走完整 pipeline", "发布这个 skill", or "skill 上线". Use when a new skill has passed review and is ready for release across one or more agent platforms.
In this repository15 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- agent-config-referenceSkill
- agent-security-guardSkill
- code-review-workflowSkill
- cross-model-verificationSkill
- git-workflow-for-agentsSkill
- incident-retro-workflowSkill
- observability-workflowSkill
- planning-workflowSkill
- requirements-clarifierSkill
- skill-authoring-workflowSkill
- skill-review-workflowSkill
- systematic-debuggingSkill
- test-design-workflowSkill
The file
Files
Skill Pipeline Orchestrator
Overview
One skill, five stages, zero manual handoffs. Author writes → Review checks → Cross-check verifies → Package bundles → Deploy installs. Each stage has a go/no-go gate. Failure at any stage loops back to fix before proceeding.
When to Use
| Use When | Don't Use When |
| 用户说"一键发布 skill""走完整 pipeline""skill 上线" | skill 还在草稿阶段(先走 skill-authoring-workflow) |
| 新 skill 审查通过后,需要打包和部署 | 只是修改已有 skill 的小问题(直接编辑,不需要走 pipeline) |
| 需要跨平台同时部署(Hermes+Claude+Codex+Cursor) | 用户说"先看看",还没决定要上线 |
| 用户说"发布这个 skill",已完成 skill-review | 审查未通过(先修完再发布) |
The Pipeline
用户触发"一键发布"
│
▼
┌─────────────┐
│ Stage 1: │
│ Author │── fail ──→ 回到 skill-authoring-workflow 修复
└──────┬──────┘
│ pass
▼
┌─────────────┐
│ Stage 2: │
│ Review + │── fail ──→ 回到 Stage 1 重写
│ Security │
└──────┬──────┘
│ pass
▼
┌─────────────┐
│ Stage 3: │
│ Cross-Check │── fail ──→ 回到 Stage 2 重审
└──────┬──────┘
│ pass
▼
┌─────────────┐
│ Stage 4: │
│ Package │── fail ──→ 回到 Stage 2 修复格式
└──────┬──────┘
│ pass
▼
┌─────────────┐
│ Stage 5: │
│ Deploy │── fail ──→ 检查平台路径,重试
└──────┬──────┘
│ pass
▼
上线完成Stage 1: Author
Loads: skill-authoring-workflow
skill-authoring-workflow 创作完成?
│
├── 是 → validate-skill.py 8/8 通过?
│ ├── 是 → Stage 1 pass → Stage 2
│ └── 否 → 修复格式 → 重跑 validate-skill.py
│
└── 否 → 加载 skill-authoring-workflow → 走完整 6 阶段Expected Output: SKILL.md 草稿 + validate-skill.py 8/8 通过
Stage 2: Review + Security
Loads: skill-review-workflow + agent-security-guard
加载 skill-review-workflow 做 23 项审查
│
├── 总分 ≥ 40?
│ ├── 是 → 加载 agent-security-guard 做安全扫描
│ │ ├── 无安全问题 → Stage 2 pass → Stage 3
│ │ └── 发现安全问题 → 修复 → 重新 Stage 2
│ │
│ └── 否 → 修复审查发现的问题 → 重新 Stage 2
│
└── 发现 Critical 级别问题 → 强制修复,不允许跳过Expected Output: skill-review 评分卡 + agent-security-guard 安全扫描报告
Stage 3: Cross-Check
Loads: cross-model-verification
用户确认要交叉验证?
│
├── 是 → 加载 cross-model-verification
│ ├── 生成对抗性提示词
│ ├── 发送第二模型
│ ├── Diff 结果
│ └── 无 Critical 发现 → Stage 3 pass → Stage 4
│ └── 有发现 → 回到 Stage 2 修复
│
└── 否(用户跳过)→ Stage 3 pass → Stage 4(跳过验证)Expected Output: 第二模型审查报告 + 分类后的发现清单
Stage 4: Package
Loads: scripts/package_skill.py
运行 package_skill.py
│
├── 验证通过 → 生成 .zip 包
│ └── Stage 4 pass → Stage 5
│
└── 验证失败 → 列出错误项
└── 回到 Stage 2 修复,重新 packagepython scripts/package_skill.py <skill-path>
# Expected: <skill-name>.zip 生成成功Expected Output: <skill-name>.zip 文件
Stage 5: Deploy
Loads: agent-config-reference(查平台路径) + scripts/install.sh
安装到各平台
│
├── 先检查目标路径是否已存在
│ ├── 已存在 → 询问用户:overwrite / backup / skip
│ └── 不存在 → 继续安装
│
├── Hermes → cp -r <skill> ~/.hermes/skills/<category>/
│ └── validate-skill.py 通过 → ✓
│
├── Claude Code → cp -r <skill> ~/.claude/skills/
│ └── validate-skill.py 通过 → ✓
│
├── Codex CLI → cp -r <skill> ~/.codex/skills/
│ └── validate-skill.py 通过 → ✓
│
└── Cursor → cp -r <skill> ~/.cursor/skills/
└── validate-skill.py 通过 → ✓全部通过 → 上线完成
Expected Output: 4 个平台路径确认 + 各平台 validate-skill.py 通过
Failure Recovery
| 失败阶段 | 失败原因 | 恢复动作 |
| Stage 1 | 格式验证失败 | 修复 frontmatter / description / 结构 |
| Stage 1 | 内容不符合类型 | 重新走 skill-authoring-workflow 分类 |
| Stage 2 | 审查分数 < 40 | 按评分卡修复项逐项修改 |
| Stage 2 | 安全扫描发现问题 | 按 agent-security-guard 报告修复 |
| Stage 3 | 第二模型发现 Critical | 回到 Stage 2 修复后重审 |
| Stage 3 | 第二模型不可用 | 降级为自审,标记 degraded,继续 |
| Stage 4 | package_skill.py 失败 | 检查文件完整性,补全 resources |
| Stage 5 | 某平台安装失败 | 检查路径和权限,手动安装 |
| Stage 5 | 目标路径已存在 | 用户选择 overwrite / backup / skip 后再继续 |
Interaction with Other Skills
- skill-authoring-workflow → Stage 1:创作阶段加载,走 6 阶段流程
- skill-review-workflow → Stage 2:23 项质量审查
- agent-security-guard → Stage 2:安全专项扫描(与 review 并行)
- cross-model-verification → Stage 3:第二模型交叉验证
- agent-config-reference → Stage 5:查各平台安装路径
- git-workflow-for-agents → Stage 5:部署后走 git 提交流程
Verification Checklist
完成前逐项确认:
- Stage 1: SKILL.md 存在 + validate-skill.py 8/8 通过
- Stage 2: skill-review 评分 ≥ 40 + agent-security-guard 扫描通过
- Stage 3: cross-model 完成(或用户确认跳过)+ 结果分类
- Stage 4: package_skill.py 成功生成 .zip
- Stage 5: 已检查目标路径;存在时已获得 overwrite / backup / skip 决策
- Stage 5: 4 个平台全部安装成功
- 任意阶段失败 → 已记录失败原因 → 已执行恢复动作
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

