Token导航 LogoToken导航TokenDH.com
开发规范external-servicegithub未标认证来源可访问clear审计未展示

inkeep-best-practices保持最佳实践

Agent Skill

inkeep-best-practices 用于补充开发规范相关能力,适合在 Codex、Claude、Cursor、Gemini CLI 中需要让 Agent 承接开发规范相关任务时使用。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

94

周安装

4

GitHub Stars

公开资料未说明

下载量

33
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

MIT

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:inkeep-best-practices(保持最佳实践)
来源仓库:https://github.com/inkeep/skills
仓库路径:skills/inkeep-best-practices
安装命令:
npx skills add inkeep/skills --skill "inkeep-best-practices"
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。

AgentSkills.tonpx skills
npx skills add inkeep/skills --skill "inkeep-best-practices"

简介

保持最佳实践技能辅助开发规范的落地执行。inkeep-best-practices 属于开发规范类 Skill,可作为该场景下的辅助能力补充。

  • 适用于代码审查、架构设计与团队协作标准化。
  • 通过 inkeep/skills 仓库安装,兼容主流 IDE 插件。
  • 使用前应导入组织内部编码规范文档作为依据。
  • 注意定期更新规则集,适应技术栈演进与新标准要求。

SKILL.md

Inkeep Agents SDK Best Practices

Comprehensive guide for building effective multi-agent systems with the Inkeep Agents SDK. Contains best practices across 7 categories, prioritized by impact to guide agent design, tool configuration, and context management.

When to Apply

Reference these guidelines when:

  • Creating new Agents or Sub Agents
  • Configuring MCP servers or function tools
  • Deciding between transfer vs delegate patterns
  • Implementing structured outputs (data/artifact/status components)
  • Managing context via headers and context fetchers
  • Organizing projects and credentials
  • Using the CLI for push/pull workflows

Rule Categories by Priority

PriorityCategoryImpactPrefix
1Sub Agent DesignCRITICALsubagent-
2Multi-Agent RelationshipsHIGHrelationship-
3Tools & MCP ServersHIGHtools-
4Context ManagementMEDIUM-HIGHcontext-
5Structured OutputsMEDIUMoutputs-
6Project OrganizationMEDIUMproject-
7Development WorkflowLOW-MEDIUMworkflow-

Quick Reference

1. Sub Agent Design (CRITICAL)

  • subagent-focused-scope - Keep sub agents focused on narrow, well-defined tasks
  • subagent-tool-limit - Limit sub agents to 5-7 related tools
  • subagent-clear-instructions - Write clear, specific instructions in prompts
  • subagent-stable-ids - Use explicit stable IDs for agents across deployments

2. Multi-Agent Relationships (HIGH)

  • relationship-transfer-vs-delegate - Use transfer for handoff, delegate for subtasks
  • relationship-delegate-no-user-response - Delegates cannot respond directly to users
  • relationship-transfer-control - Transfers make receiving agent the primary driver
  • relationship-minimal-complexity - Avoid multi-agent when single agent suffices

3. Tools & MCP Servers (HIGH)

  • tools-mcp-server-config - Configure MCP servers with proper credentials
  • tools-selected-tools - Use selectedTools to limit exposed tools
  • tools-function-tools - Use function tools for custom sandboxed logic
  • tools-credential-references - Reference credentials by ID, resolve at push time

4. Context Management (MEDIUM-HIGH)

  • context-headers - Pass request-time context via headers
  • context-fetchers - Auto-fetch external data at conversation start
  • context-variables - Reference headers/context as {{variables}} in prompts
  • context-artifacts - Save tool call results as artifacts for sharing

5. Structured Outputs (MEDIUM)

  • outputs-data-components - Use data components for rich UI rendering
  • outputs-artifact-components - Define artifact structure with props schema
  • outputs-status-updates - Provide real-time progress during long operations
  • outputs-zod-schemas - Use Zod schemas for type-safe component props

6. Project Organization (MEDIUM)

  • project-logical-grouping - Group related agents/tools in projects
  • project-model-inheritance - Configure models at project level for inheritance
  • project-credential-management - Manage credentials per environment
  • project-stop-conditions - Set stopWhen limits at project/agent level

7. Development Workflow (LOW-MEDIUM)

  • workflow-push-pull - Use CLI push/pull to sync code with Visual Builder
  • workflow-environment-settings - Use createEnvironmentSettings for env-specific config
  • workflow-minimal-config - Don't add unnecessary properties or fake MCPs

How to Use

Read individual rule files for detailed explanations and code examples:

rules/subagent-focused-scope.md
rules/relationship-transfer-vs-delegate.md
rules/tools-mcp-server-config.md

Each rule file contains:

  • Brief explanation of why it matters
  • Incorrect code example with explanation
  • Correct code example with explanation
  • Additional context and references

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

04

需要参考平台分布和安装热度时

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

补充不同宿主或平台的使用分布数据

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

Cursor

62.9%
按下载量换算21

Claude Code

33.58%
按下载量换算11

安全审计

暂无安全审计结果可展示。

权限和风险

external-service

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。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