| name | huawei-cloud-skill-creator |
| version | 2.1.0 |
| description | 1. Interactive requirements gathering via Socratic questioning — one question at a time, 4 service dimensions
2. Scaffolds complete Skill directory structure with references/, scripts/, templates/, i18n/
3. Guides through API research, dataflow diagram generation, and content creation steps
4. Validates quality with validate-skill.sh and CLI functional testing (test-cli-commands.sh)
5. Supports resource lifecycle testing for Skills with Create/Delete/Update operations
Triggers include: "创建华为云Skill","新建华为云Skill","华为云skill创建器","创建 Skill","新建 Skill","skill 创建器","create skill","build skill","new skill","skill creator","scaffold a Huawei Cloud skill","wrap CLI or OpenAPI into a skill","package cloud operations into a skill","帮我创建华为云Skill","帮我新建一个Skill","封装华为云CLI为Skill","华为云Skill脚手架","帮我创建一个skill","我需要一个skill".
|
| tags | ["huawei-cloud","skill-creator","skill-development","cli","devops","interactive"] |
Huawei Cloud Skill Creator
⚠️ MANDATORY RULE: You MUST follow the interactive questioning process before creating any files.
Ask ONE question at a time. Wait for user response. Present summary every 5 questions.
See references/usage-guide.md for detailed questioning guide.
概述 / Overview
Interactive Skill Creator for Huawei Cloud. Gathers requirements via Socratic questioning (4 dimensions: target service, functionality scope, CLI operations, trigger scenarios), then scaffolds the complete Skill directory with references, scripts, templates, i18n, and validation. Wraps Huawei Cloud CLI/OpenAPI into reusable AI Shell Skills, covering the full pipeline from requirements analysis to functional testing.
前置条件 / Prerequisites
- hcloud CLI installed and authenticated
hcloud --version
hcloud configure list
- Node.js + npx available
- Basic knowledge of the target Huawei Cloud service (ECS, VPC, OBS, etc.)
- Huawei Cloud AK/SK credentials (for resource lifecycle testing)
工作流 / Workflow
The creation process follows 4 high-level phases. See references/usage-guide.md for the full 11-step workflow.
Requirements (Interactive) → Scaffolding → Validation → Testing
Phase 1: Requirements Analysis (MANDATORY — interactive)
Socratic questioning covering 4 dimensions — one question at a time:
- Target service — Which Huawei Cloud service? (ECS, VPC, OBS, RDS, etc.)
- Function scope — What should the Skill do? (query, diagnose, deploy, monitor)
- CLI operations — Which operations? (List, Show, Create, Delete, etc.)
- Trigger scenarios — When will Agent use it? (daily inspection, troubleshooting, auto-scaling)
After every 5 questions or when all dimensions covered → present summary table → wait for user confirmation.
🛑 Do NOT proceed until user explicitly confirms.
Phase 2: Scaffolding
- Name the Skill:
{platform}-{product}-{function} (e.g., huawei-cloud-ecs-diagnosis-workflow)
- Determine domain directory (compute / network / storage / database / security / monitoring / middleware / devtools / solution)
- Create directory structure with SKILL.md, references/, scripts/, templates/, i18n/
- Generate Mermaid dataflow diagram
- Write SKILL.md and references (iam-policies, quality-checklist, etc.)
- Generate scripts (validate-skill.sh, test-cli-commands.sh) and i18n translations
Phase 3: Validation
bash scripts/validate-skill.sh {skill-path} --phase all-install
Validates: SKILL.md structure, frontmatter, required sections, dependencies, i18n, security patterns.
Phase 4: Functional Testing (Mandatory: Actual Execution + CLI→SDK→API Fallback)
每条写入 SKILL.md 的指令必须实际执行验证，按 CLI → SDK → API 三级降级：
CLI 执行示例：
hcloud ECS ListFlavors --cli-region=cn-north-4 --limit=1
| 优先级 | 执行方式 | 失败降级条件 |
|---|
| 1st | CLI — hcloud 命令 | 命令不存在或参数错误 → 先检查语法修复；修复后仍失败 → 降级 |
| 2nd | SDK — Python huaweicloudsdk | SDK 未安装或模块缺失 → 自动安装后重试；仍失败 → 降级 |
| 3rd | API — curl + AK/SK v2 签名 | 网络错误或 endpoint 不可达 → 报错并建议手动验证 |
执行流程：
写新指令 → 尝试 CLI 执行
├── ✅ 成功 → 写入 SKILL.md，继续下一条
└── ❌ 失败 → 检查是否语法问题
├── ✅ 语法问题 → 修复命令后重试 CLI
└── ❌ 非语法问题 → 降级到 SDK
├── ✅ 成功 → 写入 SKILL.md + 备注 "(SDK verified)"
└── ❌ 失败 → 降级到 API
├── ✅ 成功 → 写入 SKILL.md + 备注 "(API verified)"
└── ❌ 失败 → 标记为 "⛔ 需人工验证"，记录失败详情
只读命令（List/Show/Get/Describe/Query）直接对真实 API 执行 --limit=1 验证。
变更命令（Create/Update/Delete）先做 --help 语法检查，然后再用 --dry-run（如果支持）或最小权限参数验证参数格式，最终执行必须经用户确认。
REST API 降级规则（强制性）
当降级到 3rd API（REST API）时，严禁 Agent 自行猜测/构造 API 路径。 REST API 的 endpoint、method、path、请求体必须由用户确认提供。
降级到 API → 用户确认
│
├── Agent 询问：确认 API 端点 → 用户提供 {HTTP方法} {endpoint}{path}
│ 示例：用户确认 "POST https://bss.cn-north-4.myhuaweicloud.com/v2/promotions/benefits/activate-coupons"
│
├── Agent 使用用户提供的 API 执行验证 → 通过 → 写入 SKILL.md + 备注 "(API verified, endpoint confirmed by user)"
│
└── Agent 使用用户提供的 API 执行验证 → 失败 → 询问用户是否需要调整端点，否则标记 ⛔ 需人工验证
执行细则：
| 场景 | Agent 行为 |
|---|
| CLI/SDK 均失败，需降级到 API | 必须问用户："请提供此操作的 REST API 端点（HTTP方法 + URL）" |
| 用户提供了端点 | 使用用户提供的端点实际执行验证 |
| 用户不确定/无法提供 | 标记为 ⛔ 需人工验证，绝不自行构造虚构的 API 路径 |
| SKILL.md 中已有 API 命令 | 标注 ⚠ 此 API 端点由用户提供，请确认其有效性 |
bash scripts/test-cli-commands.sh {skill-path} --executor cli
bash scripts/test-cli-commands.sh {skill-path} --executor sdk
bash scripts/test-cli-commands.sh {skill-path} --executor api
核心命令 / Core Commands
| Command | Purpose |
|---|
bash scripts/validate-skill.sh {path} --phase all-install | Full installation & structure validation |
bash scripts/validate-skill.sh {path} --phase i18n | i18n directory validation |
bash scripts/test-cli-commands.sh {path} [--region <region>] | CLI functional test + test report |
bash scripts/generate-dataflow-diagram.sh {path} [--output <path>] | Generate Mermaid dataflow diagram |
参数确认 / Parameters
| Parameter | Required | Description | Example |
|---|
{skill-path} | Yes | Target Skill directory path | ./skills/compute/huawei-cloud-ecs-manage |
{region} | No | Huawei Cloud region | cn-north-4 (default) |
--output | No | Output path for generated files | ./references/dataflow-diagram.md |
KooCLI命令格式标准 / KooCLI Command Format
hcloud <Service> <Operation> --cli-region=<region> [--key=value ...]
| Feature | Description | Example |
|---|
| Service name | Uppercase PascalCase | ECS, VPC, IAM |
| Operation name | PascalCase | ListServers, ShowServer |
| Region param | --cli-region=<value> | --cli-region=cn-north-4 |
| Simple param | --key=value | --server_id=xxx |
| Indexed param | --key.1=value1 | --servers.1.id=xxx |
| Nested param | --key.sub_key=value | --config.protocol=vnc |
See references/cli-installation-guide.md for full CLI setup guide.
设计原则 / Design Principles
- Concise SKILL.md, verbose in references — SKILL.md body must be concise (< 250 lines). Detailed step-by-step instructions, edge-case tables, and long examples go into
references/ files. Link to them from SKILL.md.
- Spec-first — All generated Skills must conform to
references/skill-spec-generic.md
- Description-driven triggering — description must include
"Triggers include:" clause
- Security first — No hardcoded AK/SK, confirm before write operations, dry-run for high-risk ops
- Cost transparency — Generated Skills must include cost estimation logic
- Least privilege — iam-policies.md provides minimum permission policy JSON
- Idempotent first — Prefer List/Show/Get read-only operations
- i18n support — Default
i18n/zh-CN/SKILL_CN.md included
- Functional testing — Structure validation → CLI command testing → resource lifecycle test
- 每条指令必须实际执行验证 — Agent 写入 SKILL.md 的每一条 hcloud 命令，在写入过程中必须用
CLI → SDK → API 三级降级策略实际调用验证，不能只写不验
- 写指令即时验证 — 每写完一条指令，立即尝试执行。执行成功 → 写入 SKILL.md；执行失败 → 先检查语法并修复，语法无问题则执行降级策略
- 三级降级策略 — 尝试顺序：CLI（hcloud 命令）→ SDK（Python huaweicloudsdk）→ API（curl + AK/SK 签名）。上一级失败且非语法问题则降级到下一级
Edge Cases / 边界情况
| Scenario | Handling |
|---|
| User says "create now" without answering questions | Remind: requirements must be gathered first. Start questioning. |
| User answers vaguely | Use concrete examples: e.g., "Do you want a one-command list of all running servers?" |
| API operation not found in CLI | Check Huawei Cloud OpenAPI docs for alternative endpoints |
| Resource creation fails during test | Analyze error (permission/quota/param), fix and retry |
| Resource release fails | Retry 3 times, if still failing tell user to clean up manually |
| User refuses resource lifecycle test | Fall back to --help syntax validation only, note "resource test skipped" in report |
验证方法 / Verification Method
Phase 3: Quality Verification
bash scripts/validate-skill.sh {skill-path} --phase all-install
Phase 4: CLI Functional Test (actual execution required)
必须按 CLI → SDK → API 三级执行验证：
bash scripts/test-cli-commands.sh {skill-path} --executor cli
bash scripts/test-cli-commands.sh {skill-path} --executor sdk
bash scripts/test-cli-commands.sh {skill-path} --executor api
验证规则：
- 每一条写入 SKILL.md 的指令都经过至少一级实际执行验证 ✅
- CLI → SDK → API 依次降级，只有当前级失败且非语法问题才降级
- 语法问题（如参数名错误）立即修复，不降级
- 最终降级到 API 仍失败 → 标记
⛔ 需人工验证
- 阅读只命令（List/Show/Get）必须有实时 API 返回数据
- 变更命令需要用户明确确认后才执行
- 报告生成在
references/test-report.md
Resource Lifecycle Test (if applicable)
Flow: Plan → Credentials → Create → Test → Release → Report
See references/usage-guide.md for details.
参考文档 / References
Authentication
| Method | Use Case | Best Practice |
|---|
| AK/SK env vars | Development | Set HUAWEI_ACCESS_KEY / HUAWEI_SECRET_KEY |
| Temp credentials | Production | Use STS temp AK/SK + SecurityToken |
| IAM role | Cloud env | Bind IAM role for automatic auth |
User-Agent
export HCLOUD_USER_AGENT=HuaweiCloud-Agent-Skills
