本文へ移動
cccskills
無料GitHub で公開

volcengine-knowledge-search

使用 ve docs search/fetch 检索火山引擎官方文档并获取全文。Use when 用户咨询火山引擎产品的概念、用法、计费规则、部署步骤、最佳实践或服务条款,或提供 www.volcengine.com/docs/、docs.volcengine.com/docs/ 链接需要阅读正文时。

インストール方法を見る

含まれるファイル(1)

  • SKILL.md8.6 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

volcengine-knowledge-search 火山引擎文档检索

通过 ve docs search 检索官方文档,通过 ve docs fetch 获取正文。请求、响应解析、正文清洗和分页由 CLI 处理;本 skill 负责场景判断、query 改写、二次检索和来源引用。

CLI 调用约定

执行 ve 前,遵循 volcengine-cli 的 调用方标识规则。每次调用都必须在进程环境中携带:

  • VE_CALLER_TYPE=ai_agent。
  • VE_CALLER_NAME:由 Agent 根据当前实际宿主环境确定,不使用固定默认值。
  • VE_SKILL_NAME:取当前执行操作的 skill 的 frontmatter name。本 skill 执行文档检索或正文读取时,该值为 volcengine-knowledge-search;引用 CLI 的调用规则不改变操作所属 skill。

该要求覆盖 --help 检查、首次检索、改写后的检索、正文读取、分页续读及重试。每次工具调用可能启动新的 shell,必须重新传入三项变量,不依赖上一次的 export。下面示例中的 <agent-id> 与 <current-volcengine-skill> 均由 Agent 根据当前环境和 skill 元数据替换,不得原样传入。这些变量仅用于调用来源标识,不改变公开文档命令无需登录的约定。

准备

首次使用时确认本机具备两个子命令:

VE_CALLER_TYPE=ai_agent VE_CALLER_NAME="<agent-id>" VE_SKILL_NAME="<current-volcengine-skill>" \
  ve docs search --help
VE_CALLER_TYPE=ai_agent VE_CALLER_NAME="<agent-id>" VE_SKILL_NAME="<current-volcengine-skill>" \
  ve docs fetch --help

若 ve 未安装或缺少子命令,加载 volcengine-cli skill,按其「Install or upgrade the ve CLI」章节安装最新版,再检查能力。文档命令访问公开服务,无需登录、AK/SK、profile 或地域配置;安装后直接使用,无需执行资源管理的身份初始化流程。

选择流程

  1. 用户提供具体文章链接:直接 fetch,读取与问题相关的正文。
  2. 用户提出问题:补全产品名和具体方向后 search,首次检索不带产品过滤。根据标题、摘要和产品编码判断相关性。
  3. 结果过宽或不相关:改写 query;只有结果中已确认目标产品的 ServiceCodes 时,才使用该编码二次检索。不要把无关命中的编码用作过滤条件。
  4. 需要完整步骤、限制条件、精确计费规则或全文总结:对相关结果的 URL 执行 fetch,按需续读;用户要求全文时,读到没有续读提示或 has_more=false。

search:检索文档

VE_CALLER_TYPE=ai_agent VE_CALLER_NAME="<agent-id>" VE_SKILL_NAME="<current-volcengine-skill>" \
  ve docs search "对象存储 TOS 如何计费" --limit 3 --output text

搜索文本是位置参数,用引号包住完整问题。用 --output text 获取可直接阅读的 Markdown:每条包含标题、纯净 URL、ServiceCodes 和正文摘要。CLI 默认输出 JSON;需要程序化处理时显式用 --output json,结果在 documents 数组中,字段为 title、url、service_codes、snippet。

参数用途
--limit N返回文档数,正整数;一般先取 3 条
--service-code CODE限定产品,编码取自已确认相关的结果;多个产品重复传此选项
--snippet-length N每条摘要的最大 Unicode 字符数,正整数;用它控制上下文体积

改写 query

语义检索需要贴近文档表述。将口语、简称或营销名补成产品全称 + 能力描述 + 具体问题,例如把「Coding Plan 多少钱」改写为「火山方舟 Coding Plan AI 编程订阅套餐如何计费」。首搜跑偏时,换全称、近义描述或功能描述再搜。

VE_CALLER_TYPE=ai_agent VE_CALLER_NAME="<agent-id>" VE_SKILL_NAME="<current-volcengine-skill>" \
  ve docs search "火山方舟 Coding Plan AI 编程订阅套餐如何计费" --limit 3 --output text

先宽搜,确认相关结果的 ServiceCodes 为 tos 后,按产品精搜:

VE_CALLER_TYPE=ai_agent VE_CALLER_NAME="<agent-id>" VE_SKILL_NAME="<current-volcengine-skill>" \
  ve docs search "跨区域复制如何配置" --limit 5 --snippet-length 300 --output text
VE_CALLER_TYPE=ai_agent VE_CALLER_NAME="<agent-id>" VE_SKILL_NAME="<current-volcengine-skill>" \
  ve docs search "对象存储 TOS 跨区域复制如何配置" --limit 3 --service-code tos --output text

多个产品编码使用 --service-code <code-a> --service-code <code-b>,每个编码单独传递。

fetch:获取正文

支持 https://www.volcengine.com/docs/... 和 https://docs.volcengine.com/docs/...。CLI 会去掉 URL 的 query 和 fragment,text 输出依次为标题、纯净 URL、当前页正文及续读提示。

VE_CALLER_TYPE=ai_agent VE_CALLER_NAME="<agent-id>" VE_SKILL_NAME="<current-volcengine-skill>" \
  ve docs fetch "https://www.volcengine.com/docs/6349/162514?lang=zh" --output text
参数用途
--start-index N从第 N 个 Unicode 字符开始,索引从 0 起
--max-length N当前页最多返回的 Unicode 字符数,正整数;默认 5000

长文末尾会提示下一页的 --start-index。用同一个 URL 和返回的索引继续读,避免漏读或重复。以下示例先取 300 字,若续读提示为 --start-index 300,再读取下一页:

VE_CALLER_TYPE=ai_agent VE_CALLER_NAME="<agent-id>" VE_SKILL_NAME="<current-volcengine-skill>" \
  ve docs fetch "https://www.volcengine.com/docs/6349/162514" --max-length 300 --output text
VE_CALLER_TYPE=ai_agent VE_CALLER_NAME="<agent-id>" VE_SKILL_NAME="<current-volcengine-skill>" \
  ve docs fetch "https://www.volcengine.com/docs/6349/162514" --start-index 300 --max-length 300 --output text

程序化分页使用 --output json,读取 content、start_index、next_start_index、remaining_characters、has_more 和 has_content。has_more=true 时用 next_start_index 续读;结束时该字段为 null。

导航页与空正文

  • /docs/{产品编号}/{文档ID} 是具体文章路径。
  • /docs/{产品编号} 是产品文档导航页,没有单篇文章正文。识别到这种链接时,按产品名和用户关心的方向 search,从结果定位具体文章再 fetch,并向用户说明链接类型。
  • has_content=false 表示没有取得正文,不能据此编造内容。检查链接是否为导航页、过期或错误地址,再用 search 定位替代文档。

结果与错误处理

  • 按证据回答:摘要足以支撑的概念问题可以直接回答;精确规则和完整步骤先取正文。部分页面不能表述为已读完整篇。
  • 引用来源:使用 CLI 返回的纯净 URL,格式为 [文档标题](URL)。优先展示最多 3 条最相关来源,每个结论都能对应到已读取的内容。
  • 控制输出:search 用 --limit 和 --snippet-length,fetch 用 --max-length 和 --start-index。CLI 不提供旧脚本的临时文件预览机制,续读依赖分页参数。
  • 正常空结果:退出码为 0 且 documents 为空时,改写 query 或放宽产品过滤;这不代表服务故障。
  • 参数错误:按对应子命令的 --help 修正参数;数量和长度必须为正数,起始索引非负。
  • 请求失败:非零退出时读取 stderr,区分网络问题和上游业务错误。超时或 DownstreamError 等暂时性故障可重试一两次;持续失败时说明失败原因,不把失败当成无结果。公开文档请求失败无需补充云账号凭据。

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Query and answer questions about Volcengine API specifications. Use when the user asks about API parameters, error codes, request methods, enum values, required fields, response structures, pagination, parameter dependencies, service capability lists, API availability, or API comparisons, even if they do not explicitly say "API". Typical intent includes checking what an action supports, which fields are required, what values a parameter accepts, why an API returns a specific error, how to pass nested or body parameters, how pagination works, what an API response contains, whether a batch operation exists, what services or versions expose an operation, or how two APIs differ. Use API Explorer data as the authoritative source and preserve constraints, examples, and caveats found in the spec. Answer in Chinese or English. When the user needs runnable SDK code, language-specific examples, or SDK configuration, hand off to volcengine-sdk-generator. When they need CLI operations, hand off to volcengine-cli.

日本語の概要は準備中です。原文の説明を表示しています。

volcengine/volcengine-skills212026年9月23日 更新

Create and manage Volcengine cloud resources using the Volcengine CLI (`ve` command). Supports all Volcengine services including ECS, VPC, CLB, RDS, Redis, and more. Trigger this skill whenever the user asks to create, query, modify, or delete cloud resources on Volcengine, mentions the `ve` command, says "volcengine CLI", or describes infrastructure tasks such as "create an ECS instance", "set up a VPC", "list security groups", "allocate an EIP". Also trigger on Chinese prompts mentioning "火山引擎" or "火山" (e.g., "火山引擎上有哪些 ECS"、"查一下我火山的云服务器"、 "火山引擎创建一个 VPC"、"火山的 Redis 实例列一下"). Also trigger when the user encounters errors from `ve` commands and needs troubleshooting help. Also use for TOS bucket/object operations, tosutil (sometimes written tosutils), Ark CLI / arkcli model discovery, inference, generation, endpoints, fine-tuning, plans, or usage, and TLS (日志服务) / volclog log search, analysis, ingest, export, or LogCollector collection, through `ve tosutil`, `ve arkcli` and `ve volclog`.

日本語の概要は準備中です。原文の説明を表示しています。

volcengine/volcengine-skills212026年9月23日 更新

Use when the user asks to query Volcengine CloudTrail audit logs, investigate login or resource operations, or manage CloudTrail trails and backfill delivery tasks. Trigger on 操作审计, 审计日志, CloudTrail, cloud_trail, trail, 跟踪, 历史补投, or backfill in a Volcengine context. Audit queries are read-only; trail and backfill mutations require an explicit preview and confirmation.

日本語の概要は準備中です。原文の説明を表示しています。

volcengine/volcengine-skills212026年9月23日 更新

火山引擎合规最佳实践助手:一是根据用户诉求(要满足的合规标准、关键词、关注的风险等级), 从火山引擎官方内置的合规包模板里推荐该开启哪些、并标出哪些已开启;二是汇总账号当前的合规 态势,把已生效规则/合规包(官方内置 + 用户自定义)的评估结果按类别(法规 / 最佳实践 / 自定义)与严重度聚合成一份合规总览报告;三是当官方基线没覆盖时,指导用户写一条 Rego 策略 作为自定义合规规则并注册评估。可在用户确认后把推荐的模板部署为合规包。Use when 用户想做「合规检查 / 合规巡检 / 安全合规 / 合规最佳实践 / 该开哪些合规规则 / 等保合规 / 我火山账号合规吗 / 有哪些不合规 / 帮我写条自定义合规规则」,或提到火山引擎「配置审计 / Config / 合规包 / conformance pack / Rego 策略」。Trigger on 火山 / 火山引擎 / volcengine 关键词叠加合规场景。部署合规包 / 注册自定义规则属写操作,需用户确认;合规报告 与资源修复严格分离。

日本語の概要は準備中です。原文の説明を表示しています。

volcengine/volcengine-skills212026年9月23日 更新

Manage Volcengine's AI-native BaaS platform (Supabase edition / AIDAP / 火山引擎 AI 原生 BaaS 平台 Supabase 版) — a Volcengine-operated distribution that differs from official Supabase. Use when the user asks to create, inspect, or manage Volcengine Supabase or AIDAP resources — workspaces, branches, computes, SQL queries, schema changes, Auth, Realtime, Edge Functions, Storage buckets, frontend static-site hosting (Pages), API keys, connection info, or TypeScript type generation — or when a Volcengine deployment selects AIDAP as its database provider. Also trigger on Chinese prompts mentioning "火山引擎 Supabase" or "火山 Supabase". Operations run through the byted-supabase-cli command-line tool (installed via `npm i -g @byted-supabase/cli`; this is NOT the official `supabase` CLI). Do NOT use it for general database discussions, non-Supabase services (RDS MySQL, Redis), or pure client-side coding unrelated to Supabase backend management.

日本語の概要は準備中です。原文の説明を表示しています。

volcengine/volcengine-skills212026年9月23日 更新

Deploy a local project directory or Git repository to Volcengine as a running, reachable cloud service. USE WHEN: deploy to Volcengine, deploy to 火山引擎/火山, deploy this repo/project, publish current code, launch the app, run it in the cloud, expose it as a service, deploy to ECS/VKE/veFaaS, run on ECS, push to VKE, deploy as serverless/FaaS, or the user wants the agent to choose a Volcengine hosting target. If the user only asks which Volcengine deployment target to choose, use `volcengine-prepare` skill first. Not for creating a single standalone resource — use `volcengine-cli` skill for that.

日本語の概要は準備中です。原文の説明を表示しています。

volcengine/volcengine-skills212026年9月23日 更新

volcengine のスキルをすべて見る

このスキルの問題を報告する