Token导航 LogoToken导航TokenDH.com
研究检索external-servicegithub未标认证来源可访问clear审计未展示

walk-and-talk边走边说话

Agent Skill

walk-and-talk 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

371

周安装

15

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:walk-and-talk(边走边说话)
来源仓库:https://github.com/pentaxis93/aiandi
仓库路径:skills/walk-and-talk
安装命令:
npx skills add pentaxis93/aiandi --skill "walk-and-talk"
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 pentaxis93/aiandi --skill "walk-and-talk"

简介

用于查找、检索和筛选相关信息。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。

  • 适合根据关键词快速定位候选结果。
  • 使用时需确认搜索范围和过滤条件。
  • 建议结合原始 README 进一步了解实现细节。
  • walk-and-talk 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

name
walk-and-talk
description
Create talking guides for insight elicitation during physical walks. Use when you need human insight that benefits from embodied cognition, reflective thinking, or extraction of tacit knowledge. Triggers: 'walk and talk', 'walking guide', 'take a walk', 'think on a walk', 'voice recording insights'.

Walk-and-Talk Skill

Purpose: Create talking guides for insight elicitation during physical walks.


When to Use

  • You need human insight that benefits from embodied cognition
  • The topic requires reflective thinking, not rapid back-and-forth
  • Sequential exploration serves better than parallel
  • You want to extract tacit knowledge the human may not know they have

Core Principle

Design for the walking human, not the reading human.

A person on a walk:

  • Reads in glances (phone screen, limited attention)
  • Can't hold complex structures in memory
  • Processes sequentially and forgets earlier sections
  • Speaks thoughts aloud (voice recording)
  • Has unlimited time to think
  • Benefits from physical movement (cognition improves)

The guide must work within these constraints.


Guide Structure

1. The Question (Top)

One clear question that serves as lens for the entire walk. This question:

  • Fits on one screen
  • Is memorable enough to hold loosely while walking
  • Frames what kind of thinking you want
  • Can be returned to if they lose the thread

Example:

"For each phase of the ceremony, what context does it actually need—no more, no less?"

2. The Walk-Through (Body)

Sequential sections that:

  • Can be read independently (don't require remembering prior sections)
  • Trigger thinking rather than require processing
  • Are concise but substantial enough to ground reflection
  • Include just enough context to orient, not so much it overwhelms

Each section pattern:

## [Section Name]

[2-3 sentences: What this is, what happens here]

[1 sentence: The specific trigger for their thinking]

[Optional: A concrete detail that grounds the abstraction]

3. Closing Nudge (Bottom)

Brief prompt to capture anything that emerged across the whole walk. Often surfaces insights that don't fit neatly into any section.


Writing Principles

Concision Over Completeness

Every word costs attention. Cut ruthlessly. If they need more context, they'll ask in the recording.

Trigger, Don't Specify

Your job is to trigger their thinking, not direct it. "What does this phase need?" beats "Consider whether this phase needs X, Y, or Z."

Ground Abstractions

One concrete example anchors understanding better than three abstract sentences.

Sequential Independence

They will forget earlier sections. Each section must stand alone. Repeat critical context if needed.

Leave Space

The human's insight is the point. Don't fill all the space with your thinking. Ask, orient, step back.


Anti-Patterns

Anti-PatternProblemFix
Complex instructionsCan't follow while walkingSimplify to single actions
Multiple questions per sectionCognitive overloadOne trigger per section
Dense paragraphsCan't read in glancesShort paragraphs, whitespace
Assuming memoryThey forgot section 2 by section 5Each section self-contained
Over-specifyingConstrains their thinkingTrigger and step back
Abstract throughoutNothing to anchor onGround with concrete examples

Output Format Awareness

The human will produce a voice recording, which becomes a transcription. Design for this:

  • Sections that invite spoken response
  • Space for tangents (they may be the insight)
  • No need for them to "answer" formally—stream of consciousness is fine

Example: Ceremony Context Optimization

Question:

For each phase of the ceremony, what context does it actually need to do its job—no more, no less?

Section pattern:

## Phase 0-pre: Remember & Inherit

This happens BEFORE the goal is known. Claude recalls identity 
(Sutras) and scans recent sessions for continuity.

Currently loads: Full Sutras read, 5 complete session files.

What does this phase actually need, given the goal isn't known yet?

Why this works:

  • States what the phase does (orients)
  • Notes goal-awareness (critical context)
  • States current behavior (grounds in reality)
  • Single trigger question (invites thinking)

The Gratitude Test

Before finalizing, ask: Would the walking human thank me for this guide?

  • Too long? They'll stop reading.
  • Too complex? They'll get frustrated.
  • Too vague? They won't know what to think about.
  • Too directive? You'll get your thoughts back, not theirs.

*The guide serves the walk. The walk serves the insight. The insight serves the work.*

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

mcpjam

28.71%
按下载量换算33

Claude Code

22.45%
按下载量换算26

windsurf

16.06%
按下载量换算19

zencoder

13.49%
按下载量换算16

crush

8.85%
按下载量换算10

cline

3.88%
按下载量换算5

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