AGENT SKILL

voice-draft

Voice-driven document drafting for the Rabbit R1. Dictate stream-of-consciousness speech; the R1 formats it into Email, Blog Post, Report, Proposal, Meeting Notes, Creative Writing, Code Comments, Cover Letter, or Tweet Thread — then saves the result to a configured folder on your local machine via OpenClaw.

License 许可
MIT
最近更新
2026-09-04
Stars
★ 0
Forks
0

01 前言 Front Matter

name
voice-draft
author
rabbit-r1-skillz
version
1.0.0
description
Voice-driven document drafting for the Rabbit R1. Dictate stream-of-consciousness speech; the R1 formats it into Email, Blog Post, Report, Proposal, Meeting Notes, Creative Writing, Code Comments, Cover Letter, or Tweet Thread — then saves the result to a configured folder on your local machine via OpenClaw.

02 资源文件 3

03 SKILL.md 正文

VoiceDraft — OpenClaw Skill Setup

VoiceDraft turns raw speech into polished, structured documents. You dictate on the R1; the R1 sends the formatted text to your local machine via the OpenClaw bridge, which writes the file to a folder of your choice.


Prerequisites

  • OpenClaw installed and running on your local machine
  • Node.js >= 18 in your PATH (the skill daemon is a Node script)
  • A writable output folder on your machine (default: ~/Documents/VoiceDrafts/)

No other dependencies are required. The skill does not call external APIs from your machine — all LLM formatting happens on the R1.


Installation

  1. Copy the voice-draft/ folder into your OpenClaw skills directory:

    cp -r voice-draft/ ~/.openclaw/skills/voice-draft/
    
  2. Install dependencies:

    cd ~/.openclaw/skills/voice-draft
    npm install
    
  3. Edit config.json to set your output folder and preferences:

    {
      "outputDir": "/Users/you/Documents/VoiceDrafts",
      "fileFormat": "markdown",
      "addTimestamp": true,
      "overwriteRevisions": true
    }
    
    Field Default Description
    outputDir ~/Documents/VoiceDrafts Where formatted drafts are written
    fileFormat "markdown" "markdown" (.md) or "text" (.txt)
    addTimestamp true Prepend YYYY-MM-DD_HHMM_ to filenames
    overwriteRevisions true Replace file on revision rather than creating a new file
  4. Start the OpenClaw daemon if it is not already running:

    openclaw start
    
  5. Verify the skill is loaded:

    openclaw skills list
    

    You should see voice-draft in the output.


How to use VoiceDraft

Picking a document type

Scroll the R1 wheel to cycle through the 9 document types before you start:

Type Best for
Email Professional correspondence
Blog Post Engaging long-form content
Report Factual, structured business writing
Proposal Persuasive pitch documents
Meeting Notes Decisions, action items, attendees
Creative Stories, essays, expressive writing
Code Comment Function/class documentation blocks
Cover Letter Job applications
Tweet Thread Numbered, punchy social posts

Dictating a draft

  1. Select your document type by scrolling.
  2. Tap the PTT button once — the LED turns blue and the mic activates.
  3. Speak freely in a stream-of-consciousness style. Do not worry about punctuation, structure, or grammar. Just talk.
  4. Long-press PTT to end dictation. The R1 applies LLM formatting (LED turns amber), then saves the file to your machine (LED turns green).
  5. The R1 announces: "Email draft complete. 340 words, 2 minute read. Saved to email_2024-06-13_1045.md."

Reading back a draft

After a draft completes (or after loading one from history), tap PTT once. The R1 reads the full formatted document aloud via TTS.

Revising a draft

After a draft completes, long-press PTT to enter revision mode (LED turns blue). Speak your revision instruction:

  • "Make it shorter"
  • "More formal tone"
  • "Add a call to action at the end"
  • "Turn the third paragraph into bullet points"
  • "Shorten the subject line"

The R1 applies the revision, updates the file on your machine, and announces the new word count.

Browsing draft history

From the idle or done screen, scroll up or down to open Draft History. Scroll to navigate the last 10 drafts. Tap PTT to load a draft — it becomes the active draft you can read back, revise, or resend to your machine.


OpenClaw skill API

The R1 app communicates via the shared storage bridge (openclaw:cmd / openclaw:resp keys).

Command: save-draft

Writes a formatted document file to outputDir.

Request params:

{
  "type":     "email",
  "label":    "Email",
  "text":     "Subject: ...\n\nDear ...",
  "words":    340,
  "ts":       1718283045000,
  "revision": 1
}

Response:

{
  "status": "done",
  "data": {
    "path":     "/Users/you/Documents/VoiceDrafts/email_2024-06-13_1045.md",
    "filename": "email_2024-06-13_1045.md"
  }
}

Error response (any command)

{
  "status": "done",
  "data": {
    "error": "Human-readable error message"
  }
}

File naming convention

Files are named using the pattern:

{type}_{YYYY-MM-DD}_{HHMM}[_rev{N}].{ext}

Examples:

  • email_2024-06-13_1045.md
  • blog-post_2024-06-13_1120.md
  • report_2024-06-13_1410_rev3.md

When overwriteRevisions is true, revision 2 and later replace the original file and the _rev{N} suffix is not appended.


Folder structure on your machine

~/Documents/VoiceDrafts/
  email_2024-06-13_1045.md
  blog-post_2024-06-13_1120.md
  report_2024-06-13_1410.md
  tweet-thread_2024-06-14_0830.md

Each markdown file includes a YAML front matter block:

---
type: email
words: 340
created: 2024-06-13T10:45:00Z
revision: 1
---

Security notes

  • The skill writes files only to outputDir. It does not execute shell commands, read existing files outside that folder, or make network requests.
  • Voice transcripts and formatted text are sent from the R1 to your machine only. They are not transmitted to any third-party service by the OpenClaw skill.
  • The R1's LLM formatting runs on-device or via Rabbit's inference infrastructure, consistent with all other R1 Creations.
  • File contents are never logged by the skill daemon beyond the normal OpenClaw audit trail.

Troubleshooting

"Could not save to machine" on the R1

  • Confirm OpenClaw daemon is running: openclaw status
  • Check that outputDir exists and is writable: ls -la ~/Documents/VoiceDrafts/
  • Review the OpenClaw log: openclaw log --tail 20

Draft saves but file is empty

  • Ensure the text field is not empty — this can happen if LLM formatting failed silently on the R1. The R1 app falls back to raw transcript text in this case.

File appears but formatting looks wrong

  • The R1 LLM applies formatting based on the document type system prompt. For best results, speak at least 3-5 sentences. Very short input may produce minimal output.

Revision replaces entire file unexpectedly

  • Set overwriteRevisions: false in config.json if you want revision history preserved as separate files.

OpenClaw response timeout

  • The save-draft command has a 15-second timeout on the R1. If your machine is slow to respond, check OpenClaw daemon health and disk I/O.

04 同仓库更多 Skills

ajohnsonnow
ajohnsonnow/r1-warren

whisper-quiz

WhisperQuiz — voice-answered trivia game for Rabbit R1. Generates questions via the R1's built-in LLM, captures spoken answers via STT, and evaluates correctness with flexible LLM-based matching.

★ 0
ajohnsonnow
ajohnsonnow/r1-warren

walk-cast

Walking podcast production skill. Receives raw audio metadata and transcript from the R1, removes long pauses, applies noise reduction and normalization, generates AI chapter titles and episode metadata via LLM, and uploads the final episode to your configured podcast RSS host (Spotify for Podcasters / Anchor, Podbean, Buzzsprout, RSS.com, or any FTP-accessible feed).

★ 0
ajohnsonnow
ajohnsonnow/r1-warren

voice-novel

Local machine skill for VoiceNovel — handles story export to disk.

★ 0
ajohnsonnow
ajohnsonnow/r1-warren

voice-guide

Environment description and navigation assistant for blind and low-vision users. Analyzes camera frames via a vision LLM, reads visible text, identifies scene categories (people/vehicles/food/hazards/exits), performs 360° scans, retrieves GPS position, and sends emergency SMS alerts — all via audio feedback with zero screen dependency.

★ 0
ajohnsonnow
ajohnsonnow/r1-warren

voice-commit

Hands-free git and CI workflow voice controller for the Rabbit R1. Speak git/CI commands and hear results — commit, push, status, PR management, test runs, and deploy, all voice-driven via OpenClaw.

★ 0
ajohnsonnow
ajohnsonnow/r1-warren

vet-vault

VA benefits assistant with DD214 OCR, eligibility analysis, and benefit summary export. Handles camera capture → OCR extraction → eligibility computation, VA deadline fetching, and PDF export for veteran appointments.

★ 0