Token导航 LogoToken导航TokenDH.com
研究检索权限需确认github未标认证来源可访问许可证需确认审计通过

explore-and-plan探索和计划

Agent Skill

explore-and-plan 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

194

周安装

8

GitHub Stars

9

下载量

63
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:explore-and-plan(探索和计划)
来源仓库:https://github.com/patrick-fu/awesome-skills
仓库路径:skills/explore-and-plan
安装命令:
npx skills add https://github.com/patrick-fu/awesome-skills --skill explore-and-plan
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/patrick-fu/awesome-skills --skill explore-and-plan

简介

explore-and-plan 将已收敛的需求转化为可执行的落地步骤清单。

  • 不重复已有探索成果,直接产出可直接交付的行动计划。
  • 适用于方向明确但缺少实施细节的产品或工程任务。
  • 若需求仍开放讨论,应先引导至头脑风暴阶段再进入规划流程。
  • 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

Explore And Plan

The job of this skill is not to explore forever. It is to produce a plan that someone can execute without needing to rediscover the intent, architecture, or missing decisions.

Check Readiness

Before writing a plan, decide whether the request is actually ready for planning.

Use this skill when the user has a chosen or nearly chosen direction and wants it turned into concrete execution steps.

If the request is still open-ended, mostly about comparing ideas, or missing core product or technical decisions, do not fake a plan. Hand off to brainstorm first, let the exploration converge, then come back and write the plan.

Some requests land in between. In that case, resolve the last important unknowns quickly, then move into planning. Do not let the conversation drift back into broad ideation once the remaining questions are small and specific.

Scope The Plan

Before breaking work into steps, check whether the request is actually one plan or several different plans hiding inside one prompt.

If it spans multiple loosely coupled surfaces, phases, or subsystems, say so and split it into clear parts or stages before detailing execution. Do not write one giant blended plan if the work would be clearer as separate tracks.

Choose a planning scope that is coherent enough for one executor to follow without constantly switching mental models.

Map The Change Surface

Before listing steps, identify what the plan will touch:

  • files, modules, or systems likely to change
  • important existing behavior that must be preserved or replaced
  • interfaces, dependencies, and ownership boundaries
  • the expected end state

This map does not need to be long, but it should anchor the plan in the real codebase or system rather than in abstractions.

Resolve Remaining Decisions

Do not push unresolved choices into the body of the plan.

If there is still an if, or, maybe, consider, possibly, depending on, or TBD that matters to execution, resolve it before presenting the plan. Ask the user only when the decision cannot be made responsibly from context.

Good plans commit to a path, state the reasoning briefly, and make exclusions explicit so the executor does not expand scope during implementation.

Write The Plan

The plan should be self-contained, concrete, and ordered.

Include the context an executor needs:

  • what exists now that matters
  • what will change
  • why key choices were made
  • what done looks like

Write steps that are specific enough to execute. A good step names the surface it changes, the action to take, non-obvious implementation details, and how to verify correctness when verification is not obvious.

Keep the order dependency-aware. Each step should set up the next one cleanly.

Format follows function. A small fix may only need a short numbered list. A larger feature may need sections, phases, or an architecture preface. Use the lightest structure that makes the plan unambiguous.

Avoid Placeholder Plans

Do not write steps that only sound concrete.

These are warning signs:

  • "add validation" without saying what is validated or where
  • "handle edge cases" without naming the edge cases
  • "update the UI accordingly" without describing the change
  • "write tests for the above" without tying tests to behavior
  • "refactor as needed" without naming the target structure
  • "do something similar in the other module" without identifying the module

If a step could not be executed by someone new to the task, it is still a placeholder.

Self-Review

Before presenting the plan, check:

  • Is the request actually ready for planning, or should it still be in brainstorm?
  • Does each requirement map to a concrete step or explicit exclusion?
  • Have I identified the real change surface instead of writing a generic plan?
  • Is there any point where the executor would still need to make a design decision?
  • Are there filler steps that sound plausible but do not actually tell someone what to do?
  • If someone follows this in order, will they arrive at the intended end state without needing more context?

If any answer is no, the plan is not done.

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

39.79%
按下载量换算25

Claude

27.38%
按下载量换算17

Cursor

18.14%
按下载量换算11

Gemini CLI

9.07%
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

权限需确认

当前来源未能明确判断权限范围,默认进入异常复核队列。

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