AGENT SKILL

talk

Start a hands-free voice conversation with Claude through the speakers and microphone. Optionally name a task to work on.

License 许可
MIT
最近更新
2026-11-14
Stars
★ 0
Forks
0

01 前言 Front Matter

name
talk
description
Start a hands-free voice conversation with Claude through the speakers and microphone. Optionally name a task to work on.
allowed-tools
[ "mcp__plugin_mac-voice-mcp_voice-mcp__speak_and_listen" ]
argument-hint
[task to work on]
disable-model-invocation
true

03 SKILL.md 正文

Let's work in voice mode. I may be away from the screen, so keep me in the loop through the speak_and_listen tool instead of waiting for me to type.

  • This is a spoken conversation. Every reply to me goes through speak_and_listen, and each transcript is my next turn. Don't fall back to text replies until I say "stop voice mode" or "I'm back".
  • Keep the conversation going while work happens. Do quick things (a lookup, a small edit) yourself. Hand anything that takes more than a minute or so (test runs, builds, multi-file changes, research) to a background agent (Agent tool with run_in_background: true), say briefly what it's doing, and go straight back to speak_and_listen. When an agent finishes, tell me the result by voice at the next turn. If background agents aren't available, do the work yourself as before. Don't narrate every small action.
  • Don't send a separate filler line ("Okay", "Let me check") just before your answer: every voice message takes a moment to start, so the answer arrives after an odd pause. Lead with the answer itself.
  • While you work yourself, I can talk to you by starting with your name ("Claude, use the other branch"). It reaches you after your next tool call. Take it into account and acknowledge it briefly by voice at the next pause. Mention this once, in your first spoken turn.
  • If a transcript is unclear, ask again by voice.
  • Confirm by voice before anything destructive, irreversible, or that costs money.
  • If I don't answer while background agents are still working, say once "Still on it, I'll tell you when it's done." (listen: false) and end your turn without asking again. When an agent reports back, tell me the result by voice.
  • If I don't answer, don't assume yes: ask once more out loud. If there's still nothing, pause and summarize on screen, and tell me to type anything or run /mac-voice-mcp:talk to pick up again (the mic is off, so don't tell me to speak).
  • Keep writing full details (code, diffs, links) on screen as usual; the voice line is the headline.
  • Stop using voice when I say "stop voice mode", "I'm back", or start typing again. To end it, say a short goodbye with speak_and_listen and listen: false (it speaks without opening the mic).

Write text_to_speak for the ear, not the screen:

  • 1–3 short sentences, under ~40 words; lead with the outcome, then one question.
  • Plain words only: no markdown, bullets, emoji, code, file paths, URLs, stack traces or tables.
  • Describe code instead of reading it ("I added a retry to the upload function"), say file names not paths ("in index.ts"), round numbers ("about two hundred ms"), spell out symbols.
  • Ask one question at a time, answerable in a few words ("Should I deploy it — yes or no?").
  • Put the details (diffs, logs, links) in your normal on-screen reply, and say so out loud.
  • Tool output (Bash stdout, file writes) is NOT the on-screen reply. Only your own assistant message text renders.
  • A spoken turn with no assistant message text shows the user nothing — never skip the on-screen reply.

If speak_and_listen says voice-mcp is not set up, stop and suggest running /mac-voice-mcp:setup.

Task: $ARGUMENTS

If a task is given above, start by briefly telling me out loud what you're about to do for it, then ask if I want to change anything. If it's empty, start by calling speak_and_listen to say voice mode is on and ask what I'd like to work on.

04 同仓库更多 Skills