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

deep-interview深度访谈

Agent Skill

deep-interview 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

524

周安装

21

GitHub Stars

27

下载量

170
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:deep-interview(深度访谈)
来源仓库:https://github.com/aviz85/claude-skills-library
仓库路径:skills/deep-interview
安装命令:
npx skills add https://github.com/aviz85/claude-skills-library --skill deep-interview
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/aviz85/claude-skills-library --skill deep-interview

简介

用于实时构建结构化知识库,通过自适应提问持续填补信息缺口。

  • 每轮循环包含提问、记录、深化与重复,知识随回答逐步累积。
  • 输出按主题组织的笔记与证据链,支持后续报告撰写与决策参考。
  • 建议在私有目录下运行,避免敏感信息泄露至公共路径。
  • deep-interview 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Deep Interview

Conduct adaptive interviews that progressively extract knowledge and build organized knowledge bases in real-time.

Core Loop

ASK -> LISTEN -> WRITE -> DEEPEN -> REPEAT

Each cycle: ask targeted questions, capture answers into files, identify gaps, go deeper. The knowledge base grows with every answer.

Process

1. Initialize

Parse the topic from user input. Determine output path:

  • If --output <path> provided: use that path
  • If inside a project with CLAUDE.md: use project root or relevant subfolder
  • Default: create ~/Documents/interviews/<topic>-<date>/

Create the output directory and an _interview-index.md file:

# Interview: <Topic>
**Date:** <today>
**Status:** In Progress
**Depth:** <shallow|medium|deep>

## Themes Discovered
(updated as interview progresses)

## Files Created
(updated as files are written)

2. Opening Round - Broad Context

Use AskUserQuestion to understand the landscape. Ask 2-3 broad questions max per call.

First call - establish scope and the interviewee's relationship to the topic:

  • What is this topic about? (if unclear)
  • What is the user's role/expertise level?
  • What's the goal of capturing this knowledge?

IMPORTANT: After EACH AskUserQuestion response, immediately write what was learned to a file before asking more questions. Never accumulate more than one round of answers without writing.

3. Adaptive Deepening

Based on answers, identify themes (3-7 major areas). For each theme:

  1. Create a file: <theme-slug>.md
  2. Ask 2-4 targeted questions about that theme using AskUserQuestion
  3. Write answers into the theme file
  4. If a theme is complex enough, create a subfolder: <theme-slug>/ and split into sub-files

Question strategy per depth:

DepthQuestions per themeTotal roundsOutput size
shallow2-33-55-10 files
medium4-66-1010-20 files
deep8-1212-2020-40 files

Default depth: medium.

4. Question Techniques

Vary question types to extract different knowledge layers:

  • What questions: facts, definitions, components
  • How questions: processes, workflows, methods
  • Why questions: reasoning, philosophy, decisions
  • When/Where questions: context, triggers, conditions
  • Who questions: stakeholders, audiences, roles
  • What if questions: edge cases, exceptions, failures
  • Rank/Priority questions: force prioritization ("pick top 3")
  • Contrast questions: "how is X different from Y?"

Tip: Use the options field in AskUserQuestion to suggest concrete answers when possible - this makes it easier for the user and surfaces assumptions to validate.

5. Writing Rules

File naming: kebab-case, descriptive. e.g., target-audience.md, pricing-strategy.md, session-1-agenda.md

File format:

# <Theme Title>

> Source: Deep Interview, <date>

## Key Points
- Point extracted from answer
- Another point

## Details
<Expanded content from follow-up questions>

## Open Questions
- Things that still need clarification

Folder creation trigger: When a theme has 3+ sub-themes, create a folder:

output/
├── _interview-index.md
├── overview.md
├── simple-theme.md
└── complex-theme/
    ├── _index.md
    ├── sub-topic-1.md
    └── sub-topic-2.md

6. Progress Tracking

After every 3 rounds of questions, show the user a brief status:

**Interview Progress:**
- Themes covered: X/Y
- Files created: N
- Current focus: <theme>
- Estimated remaining: ~Z more rounds

7. Synthesis & Closing

When all themes are covered (or user signals done):

  1. Update _interview-index.md with final table of contents
  2. Create _summary.md - a concise synthesis of everything learned
  3. Identify and list gaps in _open-questions.md
  4. Tell user: total files created, folder structure, and suggested next steps

AskUserQuestion Best Practices

  • Max 2-3 questions per call (don't overwhelm)
  • Always provide concrete options when possible (user can still type "Other")
  • Use multiSelect: true for "which of these apply?" questions
  • Keep headers short (max 12 chars) - they show as chips/tags
  • Phrase options as clear, distinct choices - not vague
  • After receiving answers, acknowledge briefly before writing + asking more
  • If user gives short answers, ask follow-ups. If detailed, move on.

Anti-Patterns

  • Don't ask 10 questions at once - max 3-4 per AskUserQuestion call
  • Don't wait until the end to write files - write incrementally
  • Don't ask yes/no questions when open-ended would yield more
  • Don't repeat questions the user already answered
  • Don't assume knowledge - always verify with the user
  • Don't create empty placeholder files - only write when there's real content

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

32.98%
按下载量换算56

Claude

30.65%
按下载量换算52

Cursor

18.99%
按下载量换算32

Gemini CLI

10.63%
按下载量换算18

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