원클릭으로 Manus에서 모든 스킬 실행

시작하기

voice-mode

스타1

포크0

업데이트2026년 5월 20일 05:04

Transcribe a local audio file.

설치

Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.

Manus에서 실행

출처

YishenTu

YishenTu/outclaw

GitHub 저장소 열기 Creator 저장소 보기

다운로드

Manus에서 실행

voice-mode

Transcribe a local audio file to text.

When To Use

The user turn contains [audio: /abs/path/file.oga]
The user turn contains [audio: /abs/path/file.mp3]
You have a local audio file path and need the transcript before you can answer

Command

Primary — call the local ag Antigravity CLI wrapper (free under the user's subscription):

zsh -ic 'ag --add-dir <PARENT_DIR_OF_AUDIO> \
  -p "Transcribe the audio verbatim. Output only the transcript, no commentary. @<ABSOLUTE_AUDIO_PATH>"'

ag is the local .zshrc wrapper for Antigravity CLI; use zsh -ic so the shell function is loaded.
<PARENT_DIR_OF_AUDIO> is the directory containing the audio file. It MUST be passed via --add-dir or the CLI's workspace sandbox will reject the path.
<ABSOLUTE_AUDIO_PATH> must be the absolute path, prefixed with @ inside the prompt string (bare paths are treated as literal text).
Stdout is the transcript. Nothing else.

Fallback — bundled helper script (REST API, consumes GEMINI_API_KEY):

Only run this if the ag CLI wrapper is missing or its invocation fails. From your agent workspace root:

node ./skills/voice-mode/scripts/transcribe.mjs <ABSOLUTE_AUDIO_PATH>

The fallback helper recognizes .oga, .ogg, .mp3, .m4a, .mp4, .aac, .wav, .flac, .aiff, and .aif. Unsupported local audio paths fail with exit code 3; do not rename arbitrary files to bypass this check.

Rules

Extract the absolute local file path from the prompt segment.
Try the primary ag CLI invocation first.
Treat the captured stdout as the transcript. Respond to the user as if they had typed the transcript verbatim — never wrap it in quotes, never describe it.
If the primary call fails:
- Surface the stderr message to the user.
- Fall back once to the bundled helper script above.
- If the fallback also fails, surface its stderr and stop. Never invent a transcript.
If the transcript is empty, reply: I couldn't make out the audio message — can you resend or type it?
Do not invent transcript content under any circumstance.

Prerequisites

Preferred: the local ag Antigravity CLI wrapper available from interactive zsh, authenticated via the user's subscription.
Fallback only: node on PATH and GEMINI_API_KEY in the environment.

이 저장소의 다른 Skills

같은 저장소

skill-creator

YishenTu/outclaw

Create or update a skill. Use when the user asks to make a new skill, or when you recognize a repeatable multi-step workflow that should become one.

2026-05-021

skill-finder

YishenTu/outclaw

Use when the user wants to discover, examine, or evaluate agent skills from the open Agent Skills ecosystem (skills.sh / Vercel Labs) as references for building their own. Triggers on phrases like "find a skill for X", "is there a skill that does Y", "show me skills about Z", "what's on skills.sh", "look up <skill-name>", "what does the <X> skill do". The skill enforces a fit-first workflow — clarify intent, search, fetch SKILL.md, assess fit against the user's actual need, and adapt via `/skill-creator`. Never auto-installs; the goal is learn-and-adapt, not install-and-use.

2026-05-021

obsidian-cli

YishenTu/outclaw

Interact with the running Obsidian app via its CLI. Use when the user wants to move/rename files (updates wikilinks atomically), query link graph (backlinks, orphans, deadends, unresolved/broken links), list/query tags or tasks, set/remove frontmatter properties, query bases, open files in Obsidian UI, run Obsidian commands, manage plugins/themes/snippets, view file history/sync versions, resolve templates, or perform any operation requiring live Obsidian state. Trigger phrases include "rename note", "move file", "find orphans", "broken links", "backlinks", "set property", "open in obsidian", "list tags", "vault info", "run command", "manage plugin", "base query", "resolve template", etc.

2026-04-291

actionbook

YishenTu/outclaw

Browser action engine. Provides up-to-date action manuals for the modern web — operate any website instantly, one tab or dozens, concurrently.

2026-04-161

voice-mode

Transcribe a local audio file to text.

When To Use

The user turn contains [audio: /abs/path/file.oga]

The user turn contains [audio: /abs/path/file.mp3]

You have a local audio file path and need the transcript before you can answer

Command

Primary — call the local ag Antigravity CLI wrapper (free under the user's subscription):

zsh -ic 'ag --add-dir <PARENT_DIR_OF_AUDIO> \ -p "Transcribe the audio verbatim. Output only the transcript, no commentary. @<ABSOLUTE_AUDIO_PATH>"'

ag is the local .zshrc wrapper for Antigravity CLI; use zsh -ic so the shell function is loaded.

<PARENT_DIR_OF_AUDIO> is the directory containing the audio file. It MUST be passed via --add-dir or the CLI's workspace sandbox will reject the path.

<ABSOLUTE_AUDIO_PATH> must be the absolute path, prefixed with @ inside the prompt string (bare paths are treated as literal text).

Stdout is the transcript. Nothing else.

Fallback — bundled helper script (REST API, consumes GEMINI_API_KEY):

Only run this if the ag CLI wrapper is missing or its invocation fails. From your agent workspace root:

node ./skills/voice-mode/scripts/transcribe.mjs <ABSOLUTE_AUDIO_PATH>

Rules

Extract the absolute local file path from the prompt segment.

Try the primary ag CLI invocation first.

Treat the captured stdout as the transcript. Respond to the user as if they had typed the transcript verbatim — never wrap it in quotes, never describe it.

If the primary call fails:

Surface the stderr message to the user.
Fall back once to the bundled helper script above.
If the fallback also fails, surface its stderr and stop. Never invent a transcript.

If the transcript is empty, reply: I couldn't make out the audio message — can you resend or type it?

Do not invent transcript content under any circumstance.

Prerequisites

Preferred: the local ag Antigravity CLI wrapper available from interactive zsh, authenticated via the user's subscription.

Fallback only: node on PATH and GEMINI_API_KEY in the environment.