Token导航 LogoToken导航TokenDH.com
研究检索需要联网github未标认证来源可访问许可证需确认审计提醒

ai-engineeringAI 工程

Agent Skill

ai-engineering 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 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。

总安装

235

周安装

10

GitHub Stars

2

下载量

82
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

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:ai-engineering(AI 工程)
来源仓库:https://github.com/f-krause/ai-engineering-skills
仓库路径:skills/ai-engineering
安装命令:
npx skills add https://github.com/f-krause/ai-engineering-skills --skill ai-engineering
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 npx skills 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。

skills.shnpx skills
npx skills add https://github.com/f-krause/ai-engineering-skills --skill ai-engineering

简介

ai-engineering 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词快速定位候选结果时使用。

  • 它能辅助 Agent 进行 AI 工程实践,包括 SDK 适配、API 验证和架构选型。
  • 使用时需结合具体任务场景选择最合适的工具链组合。
  • 安装前建议核对来源仓库维护状态,并留意是否会触发联网、命令执行或文件读写。
  • 适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

SKILL.md

AI Engineering

This skill is the default entrypoint for AI changes in any codebase using it. Keep it loaded for any LLM call, agent, workflow, or structured AI schema change, then read only the internal subskill file that matches the task.

Always Apply

  1. Verify current SDK and provider APIs before coding.
  2. Define success before increasing complexity. Start with a minimal eval set from real tasks or failures, then optimize against that instead of intuition.
  3. Prefer the simplest architecture that meets the requirement: direct LLM call, then structured workflow, then a single agent, then multi-agent only if proven necessary.
  4. Make machine-consumed outputs schema-first. If downstream code branches on it, encode that branch in Zod for TypeScript or Pydantic for Python instead of prose. Field order matters.
  5. Separate instructions from user input, retrieved content, and tool output. Treat all non-system text as untrusted data.
  6. Version prompts, schemas, tools, and model choices together, and log enough metadata to tie behavior changes back to traces and evals.
  7. For knowledge access, do not default to vector-database RAG. First ask whether the task is better served by long context, a search tool, hybrid retrieval, or document-scoped lookup.
  8. Default to the provider already established by the project. If the project is greenfield and the task does not require another provider, OpenAI is a reasonable starting default.
  9. If a workflow sends many near-identical requests over the same long file or shared context, check provider-side prompt or context caching before inventing custom memoization.
  10. If AI behavior changes, run the project's AI-focused regression tests if they exist.

Routing

flowchart TD
    A[AI task] --> B{What changed?}
    B -->|Prompt or model call| C[subskills/llm-calls.md]
    B -->|Repeated long-context reuse or caching| I[subskills/caching/*.md]
    B -->|Structured output or tool schema| D[subskills/schema-design.md]
    B -->|Retrieval, file search, RAG, grounding| J[subskills/retrieval.md]
    B -->|Citations, evidence spans, page locations, highlights| K[subskills/citations.md]
    B -->|Confidence, abstention, review thresholds| L[subskills/confidence.md]
    B -->|Predefined multi-step flow| E[subskills/workflows.md]
    B -->|Autonomous tool loop| F[subskills/agents.md]
    B -->|Tracing and observability| G[subskills/tracing.md]
    B -->|Security, guardrails, evals| H[subskills/safety-evals.md]

Choose The Matching Subskill

  • subskills/llm-calls.md: single calls, prompt shape, retries, rate limits, model selection, and provider caching. If the issue is repeated long-context reuse, then read the matching file under subskills/caching/.
  • subskills/schema-design.md: the model contract itself, including field order, enums, discriminated unions, and lean versus audit schemas.
  • subskills/retrieval.md: how evidence is found, scoped, chunked, reranked, and passed into grounded answering.
  • subskills/citations.md: how answers map back to exact evidence locations, pages, offsets, and highlights.
  • subskills/confidence.md: how evidence quality, contradictions, and review thresholds map to confidence, abstention, and review routing.
  • subskills/workflows.md: code-owned multi-step orchestration.
  • subskills/agents.md: model-owned next-action selection inside a bounded loop.
  • subskills/tracing.md: observability, trace search, and turning executions into eval/debug inputs.
  • subskills/safety-evals.md: prompt-injection defense, guardrails, evals, and calibration.

Common combinations:

  • grounded answering: retrieval.md plus citations.md
  • grounded answering with review routing: retrieval.md plus citations.md plus confidence.md
  • any structured AI output that feeds code: schema-design.md plus the domain subskill above it

Companion Skills

Load adjacent skills when your environment includes them:

  • AI SDK or provider-wrapper skills for framework-specific invocation details
  • UI or chat-element skills for frontend rendering of model output
  • Framework-specific app skills when AI behavior is embedded in a larger web application

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

36.12%
按下载量换算30

Claude

32.14%
按下载量换算26

Cursor

17.46%
按下载量换算14

Gemini CLI

9.7%
按下载量换算8

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills