菜单
Speak a short summary aloud using Kokoro TTS. Use proactively at the end of any completed task to announce what was done in one sentence. Also use when the user explicitly asks to hear something spoken.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/dropbox/nspeech --skill speek命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Transcribe speech from the microphone. Use when the user asks you to listen, hear, or transcribe what they say. Returns the spoken text on stdout.
Transcribe speech from the microphone. Use when the user asks you to listen, hear, or transcribe what they say. Returns the spoken text on stdout.
Speak a short summary aloud using Kokoro TTS. Use proactively at the end of any completed task to announce what was done in one sentence. Also use when the user explicitly asks to hear something spoken.
基于 SOC 职业分类
正在显示 SKILL.md
| name | speek |
| description | Speak a short summary aloud using Kokoro TTS. Use proactively at the end of any completed task to announce what was done in one sentence. Also use when the user explicitly asks to hear something spoken. |
Speak text aloud using the speek command (GPU-accelerated Kokoro TTS).
In this environment, speek is approved to run outside the sandbox via the
persisted command prefix ["speek"]; call it directly so it can access GPU and
audio devices.
speek "Done. The encoder now uses private Metal buffers with blit staging."
Keep the text short (one sentence, under 20 words). Speek loads a 134MB model on first invocation (~1s), so don't call it repeatedly in quick succession.