Gerador de Cotação de Migração para Volcengine
YtYEven/volcengine-pricing-quote-skills
Sincroniza preços Volcengine e gera cotações Excel para migrações cloud
Ver no GitHubSkill autonomous que sincroniza automaticamente a tabela de preços de produtos Volcengine e gera planilhas de cotação comparativa para migrações cloud a partir de inventários Aliyun/Tencent/AWS. Inclui mapeamento de 80+ produtos, conversão de especificações e múltiplas estratégias de layout de relatório.
Instalação
npx convez-ai YtYEven/volcengine-pricing-quote-skills1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Preparar cotações comparativas para clientes migrando de outras clouds
- Manter tabela de preços Volcengine atualizada automaticamente
- Mapear especificações ECS/RDS entre provedores (AWS, Aliyun, Tencent)
Veredito

Skill especializado e bem-documentado para equipes de pré-venda e arquitetura cloud que trabalham com Volcengine. Requer ambiente Python, automação de browser e experiência com integração de skills; não é ferramenta casual, mas oferece automação robusta para fluxo repetitivo de cotações.
README
A tradução para o português deste README ainda está na fila.
火山引擎刊例价与云迁移报价 Skills
一个通用 Agent 技能包(自包含),用于:(1) 自动同步火山引擎全量产品刊例价,(2) 根据友商云清单或客户需求生成火山引擎对标报价 Excel。
English · 中文
适用的 Agent
技能包用"能力描述"而非具体工具名编写,任何满足下文"前置能力"的 Agent 框架都能跑。已验证或预期兼容:
| Agent | 状态 | 备注 |
|---|---|---|
| Claude Code (Anthropic) | ✅ 已验证 | 原生开发环境,所有功能开箱即用 |
| Codex (OpenAI) | 🟢 兼容 | 需配 Playwright MCP + Python 环境;能力命名替换见下文 |
| Trae (字节跳动 IDE) | 🟢 兼容 | 需 IDE 内置浏览器/终端能力 + Python |
| OpenClaw | 🟢 兼容 | REPL 模式下需手动确认替代 AskUserQuestion 类交互 |
| Cursor / Continue / Cline | 🟢 兼容 | 需 MCP Playwright 或同等浏览器自动化 |
| 自定义 Agent (LangChain / AutoGen 等) | 🟢 兼容 | 工具链全配齐即可 |
主流程文档(SKILL.md / sheet_strategies.md / pricing_lookup.md 等)用"建立任务清单""向用户提单选确认问题""浏览器导航"这类能力描述,每个 SKILL.md 末尾附"Claude Code 工具名对照表"作过渡参考。
前置能力(必须)
| 能力 | 哪个 skill 用 | 等价工具示例 |
|---|---|---|
| 文件读/写/编辑 | 两个都用 | Claude Code Read/Write/Edit;Codex 文件工具;Trae IDE 文件 API |
| Shell / Bash 执行 | 两个都用 | Bash 工具(用于 ls/mv/rm/python 跑 verify 脚本) |
| Python 3 环境 | 报价 skill + 验收脚本 | 系统 python3 + openpyxl 包 |
| 浏览器自动化(导航/JS evaluate/触发下载) | 仅刊例价 skill | Playwright MCP / Puppeteer / 等同能力 |
| 任务清单 / 进度追踪 | 推荐 | Claude Code TodoWrite;Codex todo;或退化到 markdown 手记 |
| 单选确认交互 | 报价 skill 推荐 | Claude Code AskUserQuestion;退化:直接问用户回答 |
最低可跑要求:文件读写 + Bash + Python 3 + openpyxl(可单独跑报价 skill,但需要手动从官网下载 CSV 喂给 Skills/火山产品刊例价/)。完整流程额外需要浏览器自动化。
包含什么
Skills/
├── 更新火山引擎产品刊例价/ # Skill 1:抓取并同步官网刊例价 CSV
│ ├── SKILL.md
│ ├── product_catalog.md # 74 个产品的代号 + 处理路径(download/snapshot)
│ ├── refresh-catalog.md # 何时/如何重建 catalog
│ ├── troubleshooting.md # 10 个已知坑 + 修复方法
│ └── scripts/verify_pricing.py
│
├── 云迁移对标报价/ # Skill 2:生成对标报价 Excel
│ ├── SKILL.md
│ ├── product_mapping.md # 阿里/腾讯/AWS → 火山 80+ 条映射,带等价度
│ ├── spec_mapping_rules.md # ECS/RDS/中间件规格映射 + 缺规格经验默认值
│ ├── pricing_lookup.md # 从本地 CSV 取价的统一规则
│ ├── sheet_strategies.md # 三种 sheet 分组策略 A/B/C + 自动推荐
│ ├── excel_template.md # 字段顺序、公式、合计、品牌分区配色
│ └── scripts/verify_quote.py
│
├── 火山产品刊例价/ # 数据源:Skill 1 写入 / Skill 2 读取的 74 份 CSV
│ └── *.csv (74 个,每个产品一份)
│
└── 火山云服务器/ # ECS 选型参考文档
├── 实例命名及选型推荐.md
└── 实例类型和规格.md
两个 Skill 怎么协作
┌─────────────────────────────────┐
│ 火山引擎官网 pricing 页面 │
└────────────┬────────────────────┘
│ (1) 浏览器自动化抓取
▼
┌──────────────────────────────────────┐
│ Skill 1: 更新火山引擎产品刊例价 │
│ (有"价格下载"按钮 → 直接下载 │
│ 无按钮 → 页面快照 + 提取表格) │
└────────────┬─────────────────────────┘
│ (2) 写入本地 CSV
▼
┌──────────────────────────────────────┐
│ 火山产品刊例价/*.csv (74 份) │
└────────────┬─────────────────────────┘
│ (3) 读取
▼
┌──────────────────────────────────────┐
│ Skill 2: 云迁移对标报价 │
│ 友商清单 → 产品对标 → 规格映射 │
│ → 取价 → 多 sheet Excel │
└────────────┬─────────────────────────┘
│ (4) 输出
▼
<客户名>-火山引擎报价-YYYYMMDD.xlsx
怎么用
1. 拉到本地
git clone https://github.com/YtYEven/volcengine-pricing-quote-skills.git
# 或把整个 Skills/ 目录拷到你的工作区任何位置
2. 装依赖
pip3 install openpyxl
# 浏览器自动化(只在跑 Skill 1 时需要):
# Claude Code: 启用 mcp__plugin_ecc_playwright
# 其他 agent: 装 playwright-mcp 或对应等价 MCP
3. 跟你的 Agent 说
| 你的需求 | 说什么 | 触发 |
|---|---|---|
| 拉取最新刊例价 | "更新一下火山引擎所有产品的刊例价" | Skill 1(全量) |
| 只更新指定产品 | "更新 ECS、RDS MySQL 的刊例价" | Skill 1(指定) |
| 阿里清单转报价 | "把这份阿里云清单做火山报价" | Skill 2(场景 1) |
| 客户直接报需求 | "客户要 N 台 4C16G ECS,给个报价" | Skill 2(场景 2) |
Agent 会读 Skills/*/SKILL.md 顶部 description 自动匹配。如果你的 agent 不会自动发现 skill,把对应 SKILL.md 内容贴进 prompt 即可。
4. 跨 Agent 移植速查
如果工具命名不同,SKILL.md 里所有"能力描述"对应的 Claude Code 工具如下,换其他 agent 时替换为等价工具:
| 能力描述(SKILL.md 用语) | Claude Code | 其他 agent 换什么 |
|---|---|---|
| "建立任务清单" | TodoWrite | Codex todo / Trae task list / 自定义 |
| "向用户提单选确认问题" | AskUserQuestion | 普通对话 + 让用户回答 |
| "浏览器导航" / "页面 evaluate" / "关闭浏览器" | mcp__plugin_ecc_playwright__browser_* | playwright-mcp / puppeteer-mcp / 自建 |
| "文件读取能力" | Read | 对应 file read 工具 |
| "Excel 写入库" | Python openpyxl | 同(无 agent 依赖) |
5. 验收脚本(强制)
两个 skill 都带验收脚本,生成后必跑、0 errors 才算完成:
# 刊例价目录验收
python3 Skills/更新火山引擎产品刊例价/scripts/verify_pricing.py
# 报价 Excel 验收
python3 Skills/云迁移对标报价/scripts/verify_quote.py <报价.xlsx> --strategy A|B|C --scenario 1|2
检查项覆盖:派生列是否为公式、合计行 SUM 范围、sheet 数量与策略匹配、品牌分区配色、CSV 同产品多份、空文件、新鲜度等。这两个脚本不依赖任何 agent,纯 Python,任何环境都能跑。
设计要点
- Agent 无关:所有主流程用能力描述,不绑死工具名。
- 自包含:整个
Skills/移到任何机器、任何 agent 都能跑,不依赖外部目录。 - 正交三层:
- Sheet 分组策略 (A/B/C) — 决定有几个 sheet
- 字段集 (4 种) — 决定每 sheet 列怎么排
- 品牌分区配色 (1 套) — 决定列怎么涂(基于列名分类,不基于位置)
- 可追溯:Excel 所有派生数都是公式(
=单价*数量),客户点 cell 能看到来源。 - 缺规格不黑盒:任何假设都在 Excel 备注列 + "假设与说明" sheet 双重声明。
- 无对标不硬塞:写"待补"+ 2~3 个候选,让客户决定。
- 自动验收:两个 verify 脚本拦机械错误(公式被改成死值、合计 SUM 范围错位、sheet 数与策略不符)。
维护
- 刊例价过时:本地 CSV mtime > 30 天 → Skill 2 启动前会自动提示先跑 Skill 1。
- 新产品上线:Skill 1 全量更新过程中若 ≥ 3 个产品代号回落到 ECS 页面 → 自动暂停 + 触发
refresh-catalog.md重建产品目录。 - 新友商产品:Skill 2 现场对标后,append 一行到
product_mapping.md,下次复用。 - 改字段顺序:改
excel_template.md+verify_quote.py同步,品牌配色靠列名自动适配,无需改色。
不在 scope 内
- 不实际下单/调用火山引擎采购 API(只产生报价 Excel)。
- 不计算"折扣价/优惠价"(只用官网刊例价 + 包1年/12 月折算)。
- 不评估业务架构是否适合迁移(架构咨询是另一回事)。
- 不抓阿里/腾讯/AWS 官网的当前刊例价(友商价不进报价单,只列产品+规格做溯源)。
License
MIT(默认选择,需要其他协议请提 issue)。
Volcengine Pricing & Cloud Migration Quoting Skills
A universal agent skill bundle (self-contained) for: (1) auto-syncing the full Volcengine product price list, and (2) generating Volcengine quote spreadsheets from competitor cloud inventories or direct customer requirements.
Supported Agents
The skill pack is written using capability descriptions rather than tool names, so any agent framework meeting the prerequisites below can run it. Verified or expected compatible:
| Agent | Status | Notes |
|---|---|---|
| Claude Code (Anthropic) | ✅ Verified | Native dev environment; works out of the box |
| Codex (OpenAI) | 🟢 Compatible | Needs Playwright MCP + Python; see capability mapping below |
| Trae (ByteDance IDE) | 🟢 Compatible | Needs IDE browser/terminal + Python |
| OpenClaw | 🟢 Compatible | In REPL mode, replace AskUserQuestion-style interactions with direct chat |
| Cursor / Continue / Cline | 🟢 Compatible | Needs MCP Playwright or equivalent browser automation |
| Custom Agents (LangChain / AutoGen / etc.) | 🟢 Compatible | As long as the toolchain is present |
Main flow documents (SKILL.md / sheet_strategies.md / pricing_lookup.md, etc.) use phrases like "create task list", "ask user a single-choice confirmation", "browser navigate" — i.e. capability descriptions. Each SKILL.md ends with a footnote mapping these to Claude Code tool names as a porting reference.
Prerequisites
| Capability | Used by | Equivalent tools |
|---|---|---|
| File read / write / edit | Both skills | Claude Code Read/Write/Edit; Codex file tools; Trae IDE file API |
| Shell / Bash execution | Both skills | Bash tool (for ls/mv/rm and running verify scripts) |
| Python 3 environment | Quote skill + verify scripts | System python3 + openpyxl package |
| Browser automation (navigate / page evaluate / download) | Only pricing skill | Playwright MCP / Puppeteer / equivalent |
| Task tracking | Recommended | Claude Code TodoWrite; Codex todo; or fallback to markdown notes |
| Single-choice confirmation | Quote skill recommended | Claude Code AskUserQuestion; fallback: ask user directly |
Minimum viable setup: file I/O + Bash + Python 3 + openpyxl (you can run the quote skill standalone but must manually download CSVs into Skills/火山产品刊例价/). Full pipeline additionally needs browser automation.
What's Included
Skills/
├── 更新火山引擎产品刊例价/ # Skill 1: scrape & sync official price list CSVs
│ ├── SKILL.md
│ ├── product_catalog.md # 74 product codes + handling path (download/snapshot)
│ ├── refresh-catalog.md # when/how to rebuild the catalog
│ ├── troubleshooting.md # 10 known pitfalls + fixes
│ └── scripts/verify_pricing.py
│
├── 云迁移对标报价/ # Skill 2: generate the quote Excel
│ ├── SKILL.md
│ ├── product_mapping.md # Aliyun/Tencent/AWS → Volcengine, 80+ entries with equivalence tier
│ ├── spec_mapping_rules.md # ECS/RDS/middleware spec mapping + defaults for missing specs
│ ├── pricing_lookup.md # unified rule to fetch prices from local CSVs
│ ├── sheet_strategies.md # three sheet-grouping strategies A/B/C with auto-recommend
│ ├── excel_template.md # field order, formulas, totals, brand-zone palette
│ └── scripts/verify_quote.py
│
├── 火山产品刊例价/ # Data source: 74 CSVs (written by Skill 1, read by Skill 2)
│ └── *.csv
│
└── 火山云服务器/ # ECS selection reference docs
├── 实例命名及选型推荐.md
└── 实例类型和规格.md
How the Two Skills Cooperate
┌─────────────────────────────────┐
│ Volcengine pricing pages │
└────────────┬────────────────────┘
│ (1) browser automation
▼
┌──────────────────────────────────────┐
│ Skill 1: Update Volcengine Price │
│ (has download button → direct CSV │
│ no button → page snapshot extract) │
└────────────┬─────────────────────────┘
│ (2) write
▼
┌──────────────────────────────────────┐
│ 火山产品刊例价/*.csv (74 files) │
└────────────┬─────────────────────────┘
│ (3) read
▼
┌──────────────────────────────────────┐
│ Skill 2: Cloud Migration Quote │
│ inventory → product map → spec map │
│ → price lookup → multi-sheet xlsx │
└────────────┬─────────────────────────┘
│ (4) output
▼
<customer>-volcengine-quote-YYYYMMDD.xlsx
Usage
1. Clone
git clone https://github.com/YtYEven/volcengine-pricing-quote-skills.git
# Or copy the entire Skills/ directory into your workspace.
2. Install Dependencies
pip3 install openpyxl
# Browser automation (only needed for Skill 1):
# Claude Code: enable mcp__plugin_ecc_playwright
# Other agents: install playwright-mcp or equivalent MCP
3. Ask Your Agent
| What you want | Say something like | Triggers |
|---|---|---|
| Refresh prices | "Update all Volcengine product prices" | Skill 1 (full) |
| Update specific products | "Update prices for ECS and RDS MySQL" | Skill 1 (targeted) |
| Aliyun inventory → quote | "Generate a Volcengine quote from this Aliyun inventory" | Skill 2 (scenario 1) |
| Customer direct requirement | "Customer wants N × 4C16G ECS, give me a quote" | Skill 2 (scenario 2) |
Agents read the Skills/*/SKILL.md description header to auto-match. If your agent doesn't auto-discover skills, paste the relevant SKILL.md into your prompt.
4. Cross-Agent Porting Cheatsheet
If your agent's tool names differ, here's the Claude Code mapping for every capability phrase used in SKILL.md:
| Capability (as used in SKILL.md) | Claude Code | Replace with on other agents |
|---|---|---|
| "create task list" | TodoWrite | Codex todo / Trae task list / custom |
| "ask user a single-choice confirmation" | AskUserQuestion | plain chat + wait for reply |
| "browser navigate" / "page evaluate" / "close browser" | mcp__plugin_ecc_playwright__browser_* | playwright-mcp / puppeteer-mcp / DIY |
| "file read capability" | Read | corresponding file-read tool |
| "Excel write library" | Python openpyxl | same (no agent dependency) |
5. Verification Scripts (Mandatory)
Both skills ship verification scripts; must pass with 0 errors before considering output done:
# Pricing directory verify
python3 Skills/更新火山引擎产品刊例价/scripts/verify_pricing.py
# Quote Excel verify
python3 Skills/云迁移对标报价/scripts/verify_quote.py <quote.xlsx> --strategy A|B|C --scenario 1|2
Checks include: derived columns are formulas, total-row SUM ranges, sheet count vs strategy, brand palette applied, duplicate CSVs, empty files, freshness, etc. These scripts are pure Python and have zero agent dependency — they run anywhere.
Design Principles
- Agent-agnostic — main flow uses capability descriptions, not specific tool names.
- Self-contained — move
Skills/to any machine, any agent; no external paths. - Orthogonal three-layer system:
- Sheet strategy (A/B/C) — how many sheets and what goes where
- Field set (4 types) — column order per sheet
- Brand palette (1 set) — column coloring (driven by column name, not position)
- Traceable — every derived value in Excel is a formula (
=unit_price*qty); click any cell to see provenance. - No black-box defaults — missing specs surface as explicit assumptions in remarks column + assumption sheet.
- No forced mapping — products without a Volcengine equivalent get "TBD" + 2~3 candidate notes.
- Auto-verify — two verify scripts catch mechanical errors (formula overwritten to literal, wrong SUM range, sheet count mismatch).
Maintenance
- Stale prices — if local CSV mtime > 30 days, Skill 2 prompts to run Skill 1 first.
- New products — Skill 1 detects ≥3 product codes falling back to ECS page → auto-pauses and triggers
refresh-catalog.mdto rebuild. - New competitor products — Skill 2 appends a row to
product_mapping.mdafter on-the-fly mapping. - Field reorder — change
excel_template.mdandverify_quote.pyin sync; brand palette adapts automatically by column name.
Out of Scope
- Does NOT actually place orders / call Volcengine procurement APIs (output is a quote Excel only).
- Does NOT compute discounted prices (uses official list prices + 1-year/12 monthly amortization).
- Does NOT evaluate whether your architecture should migrate (architecture consulting is separate).
- Does NOT scrape competitor (Aliyun/Tencent/AWS) live prices — those don't enter the quote; only product + spec are listed for traceability.
License
MIT (safe default — open an issue if you'd prefer another).