Back to Details

video-to-skill

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Video to Skill

视频转技能

Watch a tutorial, demo, or walkthrough video and generate a Claude Code skill from it.
观看教程、演示或操作讲解视频，从中生成Claude Code技能。

What This Skill Does

技能功能

Given a video demonstrating a workflow, tool, or technique, this skill:
  1. Analyzes the video to understand what's being demonstrated (via 
    eyeroll watch
    )
  2. Extracts: commands run, tools used, patterns followed, workflow steps
  3. Identifies the core capability being shown
  4. Generates a SKILL.md following the agent skills specification
  5. Optionally scaffolds implementation scripts
给定一个展示工作流、工具或技术的视频，本技能会执行以下操作：
  1. 分析视频以理解展示的内容（通过
    eyeroll watch
    ）
  2. 提取内容：运行的命令、使用的工具、遵循的模式、工作流步骤
  3. 识别视频展示的核心能力
  4. 遵循Agent技能规范生成SKILL.md
  5. 可选择生成实现脚本的脚手架

Setup

安装配置

bash
pip install eyeroll
eyeroll init          # set up Gemini API key
brew install yt-dlp   # for URL downloads
Or: 
export GEMINI_API_KEY=your-key
bash
pip install eyeroll
eyeroll init          # set up Gemini API key
brew install yt-dlp   # for URL downloads
或者：
export GEMINI_API_KEY=your-key

When To Use This Skill

适用场景

  • User shares a video and says "create a skill from this"
  • User shares a tutorial and wants to automate the workflow shown
  • User says "watch this and make it into a skill/plugin/command"
  • User shares a demo of a tool and wants a Claude Code skill that replicates it
  • User wants to turn a screen recording of a manual process into an automated skill
  • 用户分享视频并表示「基于这个创建一个技能」
  • 用户分享一个教程，想要自动化视频中展示的工作流
  • 用户说「观看这个内容，把它做成一个技能/插件/命令」
  • 用户分享一个工具演示，想要一个可以复刻该功能的Claude Code技能
  • 用户想要把手动操作的录屏转换成自动化技能

Workflow

工作流

1. Run: eyeroll watch <source> --context "create a skill from this" --verbose
2. Read the structured notes — identify:
   - What workflow/process is demonstrated
   - What commands/tools are used
   - What inputs does it take
   - What outputs does it produce
   - What are the trigger phrases (when should this skill activate)
3. Design the skill:
   - Name: concise, lowercase, hyphenated
   - Description: what it does and when to trigger
   - Instructions: step-by-step agent instructions
   - Setup: any dependencies or env vars needed
4. Generate SKILL.md following the agent skills spec
5. If the skill needs scripts, scaffold them
6. Validate: does the SKILL.md make sense? Are instructions clear?
1. Run: eyeroll watch <source> --context "create a skill from this" --verbose
2. Read the structured notes — identify:
   - What workflow/process is demonstrated
   - What commands/tools are used
   - What inputs does it take
   - What outputs does it produce
   - What are the trigger phrases (when should this skill activate)
3. Design the skill:
   - Name: concise, lowercase, hyphenated
   - Description: what it does and when to trigger
   - Instructions: step-by-step agent instructions
   - Setup: any dependencies or env vars needed
4. Generate SKILL.md following the agent skills spec
5. If the skill needs scripts, scaffold them
6. Validate: does the SKILL.md make sense? Are instructions clear?

Example Interactions

交互示例

User: "Watch this video on how to deploy to Fly.io and make a skill" Steps:
  1. eyeroll watch <url> --context "create a deployment skill for Fly.io"
  2. Notes show: user runs 
    fly launch
    , configures settings, runs 
    fly deploy
    , checks status
  3. Generate skill:
    • Name: 
      fly-deploy
    • Trigger: user asks to deploy, mentions Fly.io
    • Instructions: check for fly.toml, run fly deploy, verify health
    • Setup: requires 
      flyctl
      CLI
  4. Write 
    skills/fly-deploy/SKILL.md
User: "Turn this Loom into a skill for our release process" Steps:
  1. eyeroll watch <loom-url> --context "create a skill for our release process"
  2. Notes show: user bumps version, updates changelog, creates tag, pushes, creates GH release
  3. Generate skill that automates the full release workflow
User: "Watch how I set up a new microservice and make it repeatable" Steps:
  1. eyeroll watch <video>
  2. Notes show: user creates directory, copies template, updates config, registers in service mesh
  3. Generate a scaffolding skill that automates new service creation
用户：「观看这个如何部署到Fly.io的视频并生成一个技能」 步骤：
  1. eyeroll watch <url> --context "create a deployment skill for Fly.io"
  2. 笔记显示：用户运行
    fly launch
    、配置设置、运行
    fly deploy
    、检查状态
  3. 生成技能：
    • 名称：
      fly-deploy
    • 触发条件：用户请求部署、提到Fly.io
    • 操作说明：检查fly.toml文件、运行fly deploy、验证服务健康状态
    • 安装配置：需要
      flyctl
      CLI
  4. 编写
    skills/fly-deploy/SKILL.md
用户：「把这个Loom视频转换成我们发布流程的技能」 步骤：
  1. eyeroll watch <loom-url> --context "create a skill for our release process"
  2. 笔记显示：用户升级版本号、更新更新日志、创建标签、推送代码、创建GitHub发布
  3. 生成可以自动化完整发布工作流的技能
用户：「观看我如何搭建新微服务的操作，把它做成可复用的流程」 步骤：
  1. eyeroll watch <video>
  2. 笔记显示：用户创建目录、复制模板、更新配置、在服务网格中注册
  3. 生成可以自动化新服务创建的脚手架技能

Rules

规则

  • Always run 
    eyeroll watch
    first to understand the full workflow before generating the skill.
  • Follow the agent skills specification for SKILL.md format: 
    • name
      : max 64 chars, lowercase, hyphens only
    • description
      : max 1024 chars, describes what AND when
    • Body: clear instructions an agent can follow
  • Keep SKILL.md under 500 lines. Move detailed reference material to separate files.
  • The generated skill should be self-contained — an agent reading only the SKILL.md should understand what to do.
  • Include example interactions showing realistic trigger phrases.
  • Include setup instructions for any dependencies.
  • If the video shows a complex multi-step process, break it into clear numbered steps.
  • Ask the user to review the generated skill before finalizing — the video interpretation may miss nuances.
  • 生成技能前必须先运行
    eyeroll watch
    以理解完整工作流
  • SKILL.md格式遵循Agent技能规范： 
    • name
      ：最多64个字符，仅支持小写字母和连字符
    • description
      ：最多1024个字符，说明技能功能和触发时机
    • 正文：Agent可遵循的清晰操作说明
  • SKILL.md文件长度不超过500行，详细参考资料移至单独文件
  • 生成的技能应该是自包含的——仅阅读SKILL.md的Agent应该能理解需要执行的操作
  • 包含展示真实触发话术的交互示例
  • 包含所有依赖的安装配置说明
  • 如果视频展示了复杂的多步骤流程，将其拆分为清晰的编号步骤
  • 最终确定前请用户审核生成的技能——视频解读可能会遗漏细节

SKILL.md Template

SKILL.md模板

markdown
---
name: {skill-name}
description: >
  {What this skill does and when to use it.}
---
markdown
---
name: {skill-name}
description: >
  {What this skill does and when to use it.}
---

{Skill Name}

{Skill Name}

{One-liner description.}
{One-liner description.}

What This Skill Does

What This Skill Does

{2-3 sentences explaining the capability.}
{2-3 sentences explaining the capability.}

Setup

Setup

{Dependencies, env vars, installation steps.}
{Dependencies, env vars, installation steps.}

When To Use This Skill

When To Use This Skill

{Bullet list of trigger conditions.}
{Bullet list of trigger conditions.}

Workflow

Workflow

{Numbered steps the agent should follow.}
{Numbered steps the agent should follow.}

Example Interactions

Example Interactions

{2-3 realistic examples with user input and expected agent actions.}
{2-3 realistic examples with user input and expected agent actions.}

Rules

Rules

{Guardrails and constraints.}
undefined
{Guardrails and constraints.}
undefined