搜索

PythonReactSEOPDF
GitHub Skill

talk

Speak about any topic using the macOS say command. Use for "talk about what happened", "talk about this document", detailed discussions, or conversational narration. Adapts tone and length to match the context and user intent.

o3-cloud☆ 0 仓库 Stars0 仓库 Forks

安装与使用

请根据来源仓库和你的 AI 工具说明安装此技能。先阅读技能文件和依赖要求,再添加到工作流。

查看来源 ↗使用指南

技能说明

Talk Skill

Speak about topics, findings, and ideas using the macOS say command. Adapts automatically—brief summaries when you need quick updates, detailed discussions for deeper understanding, conversational narration for thinking through problems together.

When to Use

Respond when you hear phrases like:

  • "Talk about what just happened" - Quick recap or update
  • "Talk about this document" - Detailed discussion or analysis
  • "Talk about the reason you suggested that" - Explain reasoning
  • "Tell me about [topic]" - Deep dive or conversational exploration
  • "Say something about..." - Open-ended discussion
  • "Walk me through..." - Step-by-step narration
  • "Explain your thinking on..." - Detailed reasoning or analysis

How It Works

  1. Identify the context - What does the user want to hear about? How deep should this go?
  2. Choose the length - Match user intent: brief summary (1-2 sentences), detailed explanation (2-4 sentences), or conversational discussion (multiple short segments)
  3. Speak it naturally - Use Bash to run say command with your response, using natural language and conversational tone

Usage

say "Your summary here"

Examples

Quick Summary:

User: "Talk about what just happened"
Claude: say "Fixed the authentication bug, ran tests, and all passed."

Detailed Discussion:

User: "Talk about this document"
Claude: say "This document outlines the template system for VibeGuard. It defines three operation modes: AI-assisted setup, direct template selection, and template discovery. The system is organized into three implementation phases—adding flag and assist integration first, then expanding templates, and finally simplifying the codebase by reducing duplication. Overall, this approach reduces over fifteen hundred lines of code by making templates the single source of truth."

Conversational Reasoning:

User: "Talk about the reason you suggested that"
Claude: say "I suggested using Redis because it handles high-throughput caching really efficiently. It's in-memory, so lookups are super fast. Plus, it works great with cloud deployments and integrates naturally with most backends."

Step-by-Step Narration:

User: "Walk me through the process"
Claude: say "First, the agent analyzes the project by running vibeguard init with assist. That generates a detailed analysis. Next, it lists templates to see what's available. Then it selects the best matching template based on the analysis. It applies that template, runs vibeguard check to validate, and if there are failures, it fixes them and re-runs checks until everything passes."

Implementation

In your Claude Code response:

  1. Assess the user's intent - Are they asking for a quick recap, deeper understanding, or conversational narration?
  2. Generate appropriate content - Match the length and depth to the context
  3. Call Bash with: say "Your response here"
  4. Keep natural pacing - Brief summaries are concise, detailed explanations are 2-4 sentences, conversations can be longer if it makes sense

Guidance:

  • Quick recaps: 1-2 sentences, under 30 words
  • Explanations: 2-4 sentences, under 100 words, conversational
  • Discussions: 3-6 sentences, natural pacing, okay to explain nuance

Prerequisites

  • macOS with built-in say command
  • System volume enabled
  • Audio output device selected

Troubleshooting

No audio?

# Check volume
osascript -e 'output volume of (get volume settings)'

# Increase volume to 50%
osascript -e 'set volume output volume 50'

# Test the say command
say "test"

Customize speech:

# Slower speech
say -r 100 "Your message"

# Faster speech
say -r 200 "Your message"

# Choose voice
say -v Oliver "Your message"

使用示例

查看来源仓库中的使用示例与配置说明。

查看来源 ↗

来源历史

收录时间
2026-01-20
索引更新时间
2026-01-26
Branch
main

收录与同步日期属于本站记录。源码修改历史请查看来源仓库。

GitHub ↗

chat Comments (0)

chat_bubble_outline

No comments yet. Be the first to share your thoughts!