Token导航 LogoToken导航TokenDH.com
开发只读github未标认证来源可访问许可证需确认审计通过

optimize-prompt优化提示

Agent Skill

用于辅助提示词、系统指令、Agent 行为约束和工作流模板的整理。它适合让 Agent 规范任务边界、统一输出格式、拆分操作步骤或优化提示词可复用性。使用时需要保留真实业务约束,不要把示例当硬规则;涉及自动执行、外部工具或高风险操作时,应在提示词中明确确认步骤、权限边界和失败处理方式。

总安装

220

周安装

9

GitHub Stars

1

下载量

71
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:optimize-prompt(优化提示)
来源仓库:https://github.com/zackbart/skills
仓库路径:skills/optimize-prompt
安装命令:
npx skills add https://github.com/zackbart/skills --skill optimize-prompt
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/zackbart/skills --skill optimize-prompt

简介

提升提示词质量与执行效率优化器。

  • 结构化分解复杂任务为原子步骤。
  • 统一输出格式与错误处理约定。optimize-prompt 属于开发类 Skill,可作为该场景下的辅助能力补充。
  • 禁止将示例当作硬性规则套用。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。
  • 高风险操作必须加入人工确认环节。

SKILL.md

Prompt Drafting Assistant

The user wants to craft an effective prompt. Their intent: $ARGUMENTS

Your Process

Step 1: Understand the Goal

Based on the user's description, determine:

  • What type of prompt is needed (system prompt, task/user prompt, agentic/tool-use prompt)
  • What the prompt needs to accomplish
  • What context or constraints exist

If the description is too vague to draft a good prompt, ask 2-3 focused clarifying questions before proceeding. Keep questions specific — do not ask open-ended "tell me more."

Step 2: Draft the Prompt

Write a complete, ready-to-use prompt based on the user's intent. Follow these principles:

Structure

  • Use XML tags to separate concerns for prompts over 15 lines
  • Put role/persona first, capabilities second, constraints third, output format last
  • Keep it minimal — every line is a constraint, unnecessary ones create unexpected behavior

Clarity

  • Specify desired output format explicitly
  • Explain *why* constraints exist (models generalize better from reasons than bare rules)
  • Use action verbs ("Analyze this code" not "Could you suggest some analysis")

Model-specific considerations The guidance below is Claude-focused. Adapt for other models where applicable — the structural principles (role first, minimal constraints, explicit format) are universal.

  • Avoid over-aggressive instructions ("ALWAYS", "CRITICAL", "MUST") — these cause overtriggering on modern models. Use normal language instead.
  • For agentic prompts: address tool usage, reversibility, state management, and when subagents are warranted

Present the draft in a code block so it is easy to copy.

Step 3: Explain Key Decisions

After the draft, briefly explain:

  • Structural choices: Why you organized it this way
  • What you included and why: 3-5 key decisions
  • What you intentionally left out: Constraints you chose not to add

Step 4: Offer Refinement

Ask the user if they want to adjust anything. Common refinements:

  • Tone/formality level
  • More or fewer constraints
  • Adding examples
  • Changing output format

How to Use This Skill

For complex prompts (over 20 lines) or when unsure about best practices, read the relevant reference file before drafting:

  • System prompts: Read references/system-prompts.md for role structuring, constraint ordering, and XML tag patterns
  • Agentic/tool-use prompts: Read references/agentic-prompts.md for tool usage, state management, and subagent patterns

Skip these for simple prompts — only consult when the added depth would meaningfully improve the draft.

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

37.81%
按下载量换算27

Claude

31.62%
按下载量换算22

Cursor

19.46%
按下载量换算14

Gemini CLI

8.19%
按下载量换算6

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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