Token导航 LogoToken导航

Agent Skill · 研究检索

aomi-build(奥米建造)

aomi-build 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

aomi-build 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为调用外部服务;许可证为 MIT-0,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 27,103。

总安装 3,445周安装 148GitHub Stars 公开资料未说明下载 1,208核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
0.1.0

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:aomi-build(奥米建造)
来源仓库:https://github.com/merkle-seeds/aomi-build
安装命令:
openclaw skills install aomi-build
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install aomi-build

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

当用户想要从 API 文档、OpenAPI 或 Swagger 规范、SDK 文档、存储库示例、端点而不是构建、搭建或更新 Aomi 应用程序/插件时使用

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install aomi-build;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要调用外部服务,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

aomi-build 怎么安装?

在终端执行 openclaw skills install aomi-build。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

aomi-build 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

aomi-build 需要什么权限?

权限级别是调用外部服务。该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

aomi-build 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

aomi-build 用的人多吗?

公开数据是累计安装 3,445 次,近一周安装 148 次,下载 1,208 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

aomi-build 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。

先想清楚的地方

  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
aomi-build

奥米建造 · 本页

3,445—调用外部服务MIT-0有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
aomi-build
description
>
compatibility
Best when a local aomi-sdk checkout is available, often at ../aomi-sdk. Falls back to bundled references when the SDK repo is not present.
license
MIT
allowed-tools
Bash
metadata
author
aomi-labs
version
0.1

Aomi Build

Use this skill for tasks like:

  • "Build an Aomi app from this OpenAPI spec."
  • "Turn these REST endpoints into an Aomi plugin."
  • "Scaffold a new Aomi SDK app for this product/API."
  • "Update an existing Aomi app to support these new endpoints."
  • "Turn these builder docs or SDK repos into an Aomi assistant."

First Read

If a local aomi-sdk checkout exists, inspect these first:

  • sdk/examples/app-template-http/src/lib.rs
  • sdk/examples/app-template-http/src/client.rs
  • sdk/examples/app-template-http/src/tool.rs
  • docs/repo-structure.md
  • docs/host-interop.md
  • 2 or 3 relevant apps under apps/*/src/{lib,client,tool}.rs

If the supplied docs mostly point to GitHub repositories, SDKs, or examples instead of listing public endpoints:

  • treat those linked repositories as the real source of truth
  • inspect their README, config examples, example commands, and RPC/API surfaces
  • check whether they expose or produce a runnable service interface such as REST, GraphQL, JSON-RPC, gRPC, webhooks, or another stable client contract
  • prefer building against that executable surface instead of wrapping the docs themselves
  • avoid inventing a public transactional API that the docs do not actually publish

If the current repo is aomi-widget, also inspect:

  • apps/landing/content/examples/*.mdx
  • apps/landing/content/guides/build/**/*.mdx

If the SDK repo is not available, read:

Default Workflow

  1. Identify the product surface:

- What external API, SDK, repo, or spec is the source of truth? - What concrete callable surface exists: REST, GraphQL, JSON-RPC, gRPC, webhook, CLI contract, or something else? - Is there a real target we can point the app at: hosted service, self-hosted node, local example stack, or customer-provided endpoint? - Is this read-only, execution-oriented, or mixed? - What auth/env vars are required? - What user state must come from the host or caller? - Is this actually a public end-user API, a standard client interface exposed by a runtime/example app, or only builder-facing documentation?

  1. Describe the intended user-facing toolset before implementation:

- list the proposed tools by name - say what user intent each tool serves - call out which tools are read-only, which prepare actions, and which write or submit - mention any expected target URL, runtime, or host dependency - if the toolset is uncertain, surface the uncertainty before coding - identify the primary user workflow the app should make easy first - keep the first pass to the smallest sufficient toolset for that workflow unless the user asked for broader API coverage

  1. Reduce the spec into semantically meaningful tools.
  2. Scaffold or update the Aomi app using the standard file split:

- lib.rs for manifest and preamble - client.rs for HTTP client, auth, models, and normalization - tool.rs for DynAomiTool implementations

  1. Write the preamble around actual tool behavior, confirmation rules, and any host handoff.
  2. Validate with the SDK build flow and add focused tests when logic is non-trivial.

Tool Design Rules

  • First decide what kind of app this should be:

- product client - execution assistant - builder / SDK / runtime assistant

  • Before implementing, state the proposed toolset in concrete user-facing terms. This is part of the design, not optional polish.
  • Prefer the smallest sufficient toolset that makes the primary user workflow work end to end.
  • If there are multiple plausible integration targets, briefly state which one you are choosing and why before coding.
  • Prefer tools that interact with an actual product surface over tools that merely restate documentation.
  • A hosted API is not required. A self-hosted service, local example stack, standard RPC server, or other runnable interface still counts as a real integration target.
  • If the source material is SDK- or architecture-heavy, first ask whether it produces a service that clients call. If yes, build the client for that service.
  • Only fall back to a builder-oriented or docs-oriented tool surface when no stable executable target is available.
  • Do not mirror every endpoint 1:1 unless that is actually the cleanest model-facing API or the user explicitly asked for broad coverage.
  • Prefer 3 to 8 tools with clear user intent boundaries such as search_*, get_*, build_*, submit_*, list_*, or resolve_*.
  • Prefer intent-shaped tool names over raw protocol or transport names when practical.
  • Aggregate noisy upstream endpoints behind a smaller tool surface when the model does not need the raw distinction.
  • Prefer typed arguments over raw JSON string blobs when the primary workflow can be modeled cleanly that way.
  • Separate core tools from escape hatches. A generic fallback tool such as *_rpc or *_raw is fine, but it should not replace a clean core workflow.
  • Keep args typed and documented with JsonSchema. Field doc comments are model-facing and matter.
  • Return stable JSON with predictable keys. Normalize upstream naming, paging, and inconsistent shapes inside client.rs or helper functions.
  • Convert upstream errors into short actionable messages. Do not leak raw HTML, secrets, or giant payload dumps.

File Responsibilities

lib.rs

  • Keep it easy to scan.
  • Define PREAMBLE or a small build_preamble() hook.
  • Register tools with dyn_aomi_app!.
  • Only keep manifest-level wiring here.

client.rs

  • Own the app struct, HTTP client, auth headers, env vars, typed models, and response normalization.
  • Prefer reqwest::blocking::Client with explicit timeouts for sync tools, matching the current SDK examples.
  • Keep third-party API quirks here instead of spreading them across tool implementations.

tool.rs

  • Implement DynAomiTool.
  • Use descriptions that tell the model when to call the tool, not just what endpoint it wraps.
  • Map normalized client results into concise JSON results.
  • Use DynToolCallCtx when host state such as connected wallet, session state, or caller attributes is needed.

Preamble Rules

Write the preamble from the app's real contract:

  • Define role, capabilities, workflow, and guardrails.
  • Mention tool order for multi-step flows.
  • State explicit confirmation requirements before write actions.
  • If dates matter, include the current date or instruct the app to use exact dates.
  • If the app relies on host wallet/signing tools, say that clearly and do not imply hidden infrastructure.

For deeper patterns and examples, read references/aomi-sdk-patterns.md.

Host Interop And Execution

For execution-oriented apps:

  • Follow the public host conventions from docs/host-interop.md.
  • Do not invent private namespaces or internal fallback behavior.
  • When the next step belongs to the host wallet or signer, return machine-readable SYSTEM_NEXT_ACTION guidance.
  • Preserve exact transaction or signature args when a downstream host tool must execute them.
  • Do not claim a write succeeded until the upstream API submit step has actually completed.

Validation

When working inside aomi-sdk:

  • Scaffold with cargo run -p xtask -- new-app <name> if starting from scratch, or copy sdk/examples/app-template-http.
  • Build the plugin with cargo run -p xtask -- build-aomi --app <name>.
  • If build-aomi reports zero built plugins for a brand new app, check whether the new apps/<name>/Cargo.toml is still untracked. The current xtask prefers git ls-files discovery for app manifests.
  • For a direct compile signal on an untracked app, use cargo build --manifest-path apps/<name>/Cargo.toml.
  • If the app has meaningful branching or normalization logic, add unit tests with aomi_sdk::testing::{TestCtxBuilder, run_tool, run_async_tool}.
  • If a real target is available, validate the app with a short ladder:

- compile/build - connectivity check - one representative read flow - one representative write or submit flow when applicable - post-write verification such as status, receipt, or refreshed state

  • Prefer proving one end-to-end user scenario over checking many disconnected endpoints.

When the task also touches docs or demos in aomi-widget, update the relevant examples or guides to match the new app behavior.

Output Expectations

Aim to leave behind:

  • a coherent Aomi app crate or patch
  • typed tool args and strong descriptions
  • a preamble that explains the tool contract and rules
  • stable JSON outputs for the host/model
  • an app that can point at a real product surface when one exists
  • a short validation pass or a clear note about what could not be verified

平台分布

OpenClaw

90.53%
按下载量换算1,094

安全审计

VirusTotal

可疑

ClawScan

通过

Static analysis

通过

权限和风险

调用外部服务

该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills