Token导航 LogoToken导航TokenDH.com
研究检索执行命令github未标认证来源可访问clear审计未展示

reasonreason 效率

Agent Skill

reason 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

333

周安装

14

GitHub Stars

公开资料未说明

下载量

116
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:reason(reason 效率)
来源仓库:https://github.com/zpankz/mcp-skillset
仓库路径:skills/reason
安装命令:
npx skills add zpankz/mcp-skillset --skill "reason"
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 zpankz/mcp-skillset --skill "reason"

简介

发现并安装 AI 代理的技能。

  • 适用于推理增强、逻辑链构建与复杂问题拆解场景。
  • 可嵌入 Claude、Codex、Cursor 或 Gemini CLI 提升分析深度。
  • 安装命令:npx skills add zpankz/mcp-skillset --skill "reason"。
  • 依赖上游技能库版本兼容性,建议同步检查依赖项状态。

SKILL.md

name
reason
description
Understand through recursive decomposition and modular reconstruction

reason

structure

graph LR
    A[parse] --> B[branch]
    B --> C[reduce]
    C --> D[ground]
    D --> E[emit]

    style A fill:#e1f5ff
    style B fill:#fff4e1
    style C fill:#ffe1f5
    style D fill:#e1ffe1
    style E fill:#f5e1ff

[[#parse]] → [[#branch]] → [[#reduce]] → [[#ground]] → [[#emit]] ^structure


parse

  1. receive input ^input
  2. extract[^2]:

- #component/parameter — what varies ^parameters - #component/constant — what is fixed ^constants - #component/operation — what transforms ^operations - #component/unknown — what is sought^unknowns

  1. A question is a function awaiting its arguments[^3].

Identify what arguments it needs. ^function-analogy


branch

Generate multiple resolution paths before committing[^4]:

Path TypeTagPurposeAnchor
Typical#path/typicalThe conventional approach^path-typical
Alternative#path/alternativeA valid but less obvious route^path-alternative
Boundary#path/boundaryThe edge case or inversion^path-boundary

Assign rough likelihood. Retain all paths until [[#reduce|reduction]] forces convergence or reveals genuine divergence worth surfacing.

[!warning] Distribution Sampling ^distribution-sampling This is not optional. Single-path reasoning collapses to the mode. The goal is to sample from the full distribution of valid approaches.

reduce

Transform step by step[^5]. Each step follows the primitive: ^transformation-primitive

$$\text{input} \xrightarrow{\text{operation}} \text{output}$$

[!caution] Skip Nothing ^skip-nothing An obvious step is still a step. A hidden step is where errors hide.

five primitive operations

OperationTagDefinitionAnchor
select#primitive/selectcondition → one of two branches^op-select
sequence#primitive/sequenceoutput of one feeds [[SKILL#^input\input]] of next^op-sequence
iterate#primitive/iterateapply operation n times^op-iterate
recurse#primitive/recursedefine in terms of self, with base case^op-recurse
search#primitive/searchfind first element satisfying constraint^op-search

Name which [[SKILL#five primitive operations|primitive]] you are applying.

Substitution is explicit:

Given X, applying rule R yields X transformed by R

Never say "therefore" without showing the step that licenses it. ^explicit-substitution


ground

Before [[#emit|emitting]], verify:

  • [ ] Each conclusion traces to stated premises only^premise-tracing
  • [ ] Terms from separate contexts are not conflated^context-separation
  • [ ] The result is terminal (no further reduction possible) or explicitly

partial (further [[SKILL#^input|input]] required) ^termination-check

convergence logic

StateTagMeaningAnchor
Converged#state/convergedconfidence is high^state-converged
Diverged#state/divergedpresent the divergence—it is the answer^state-diverged
Continuing#state/continuing[[#reduce\reduction]] could continue^state-continuing
Blocked#state/blockedstate what is missing^state-blocked

emit

Output the synthesis.

ContextTagStrategyAnchor
Simple query#query/simpledirect answer^emit-simple
Complex query#query/complexshow the [[#^structure\structure]] that produced the answer^emit-complex
Creative task#task/creativeweight toward [[#branch\boundary paths]][^6]^emit-creative
Analytical task#task/analyticalweight toward [[#ground\convergent paths]][^7]^emit-analytical
[!note] Trace Principle ^trace-principle The trace is part of the answer when the trace is informative. The trace is omitted when it would obscure.

[^1]: Recursive decomposition breaks complex problems into self-similar subproblems until base cases are reached. [^2]: Component extraction as defined in [[#parse|parse phase]] identifies [[SKILL#^parameters|parameters]], [[SKILL#^constants|constants]], [[SKILL#^operations|operations]], and [[SKILL#^unknowns|unknowns]]. [^3]: See [[SKILL#^function-analogy|function analogy]] for the functional interpretation of queries. [^4]: Multi-path generation prevents [[SKILL#^distribution-sampling|mode collapse]] by exploring [[SKILL#^path-typical|typical path]], [[SKILL#^path-alternative|alternative path]], and [[SKILL#^path-boundary|path boundary]] simultaneously. [^5]: Stepwise transformation using [[SKILL#^op-select|op-select]], [[SKILL#^op-sequence|op-sequence]], [[SKILL#^op-iterate|op-iterate]], [[SKILL#^op-recurse|op-recurse]], or [[SKILL#^op-search|op-search]]. [^6]: Boundary paths (see [[SKILL#^path-boundary|path-boundary]]) explore inversions and edge cases suitable for creative divergence. [^7]: Convergent paths (see [[SKILL#^state-converged|state-converged]]) prioritize analytical rigor through [[SKILL#^premise-tracing|premise-tracing]] and [[SKILL#^context-separation|context-separation]].

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

OpenCode

30.68%
按下载量换算36

Claude Code

24.78%
按下载量换算29

windsurf

19.5%
按下载量换算23

Codex

12.1%
按下载量换算14

kiro-cli

7.17%
按下载量换算8

mcpjam

3.55%
按下载量换算4

安全审计

暂无安全审计结果可展示。

权限和风险

执行命令

安装流程涉及命令执行,可能通过 npx skills add zpankz/mcp-skillset --skill "reason" 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

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