Token导航 LogoToken导航

Agent Skill · 研究检索

openclaw-voice-patch(OpenClaw voice patch 搜索)

用于辅助音频、音乐、语音转写、语音合成或声音素材处理。它适合让 Agent 生成配乐说明、整理音频流程、调用语音工具或处理播客和视频配音素材。使用时需要确认输入音频来源、输出格式、时长和模型限制;涉及人声克隆、版权音乐或公开发布时,应先核对授权和合规边界。

openclaw-voice-patch 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为操作浏览器;许可证为 MIT-0,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 25,248。

总安装 3,816周安装 159GitHub Stars 公开资料未说明下载 1,272核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.0.1

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:openclaw-voice-patch(OpenClaw voice patch 搜索)
来源仓库:https://github.com/elzershark/openclaw-voice-patch
安装命令:
openclaw skills install openclaw-voice-patch
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install openclaw-voice-patch

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

修补 OpenClaw Control UI 以添加双语音输入按钮。适用于自动发送和连续语音输入功能扩展。支持 OpenClaw 更新后的界面适配和功能增强。安装命令:openclaw skills install openclaw-voice-patch。需确认权限范围和维护状态,避免触发 UI 修改或系统重启。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install openclaw-voice-patch;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要操作浏览器,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

openclaw-voice-patch 怎么安装?

在终端执行 openclaw skills install openclaw-voice-patch。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

openclaw-voice-patch 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

openclaw-voice-patch 需要什么权限?

权限级别是操作浏览器。该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

openclaw-voice-patch 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

openclaw-voice-patch 用的人多吗?

公开数据是累计安装 3,816 次,近一周安装 159 次,下载 1,272 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

openclaw-voice-patch 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。

先想清楚的地方

  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
openclaw-voice-patch

OpenClaw voice patch 搜索 · 本页

3,816—操作浏览器MIT-0有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
openclaw-voice-patch
description
Patch OpenClaw Control UI to add hold-for-continuous voice input (single button: short click = auto-send, hold 3s = continuous mode). Use when OpenClaw voice input needs patching, after OpenClaw updates (patches are overwritten by every update and must be re-applied), or when asked about voice input fixes. If the agent cannot access the OpenClaw installation files (permission denied or outside workspace), it will ask the user to copy the files into the workspace for patching, then copy them back.

OpenClaw Voice Input Patch

Adds smart voice input to the Control UI chat — same one mic button, but with hold-to-activate.

What This Does

Before: One mic button. Speak → text appears → must press Enter manually.

After: Same one mic button — but smarter, with hold-to-activate:

  • Short click → Auto-send: mic activates, speak, on pause text is sent automatically, mic turns off. If you say nothing for ~3 seconds, the mic auto-stops (browser SpeechRecognition timeout) — acts as a natural "misclick" safeguard.
  • Hold 3 seconds → Continuous mode: mic stays on after each message. After every speech pause, text is sent and recognition restarts. Click again to stop.
  • Live interim transcript shown above the input field while either mode is active.
  • Tooltip adapts: "Voice (hold 3s for continuous)" when idle, "Stop voice auto" or "Stop voice continuous" when recording.

Note: If the browser shows "microphone denied" errors, the microphone=(self) Permissions-Policy fix is also needed. This was merged into OpenClaw on April 18, 2026 — if you're on version 2026.4.18+, it's already included. On older versions, see the Troubleshooting section.

Step 1: Find the file

There is exactly one file to patch. It's always at:

.../openclaw/dist/control-ui/assets/index-<hash>.js

The hash in the filename changes with every OpenClaw version. Find it with:

Linux / Mac:

find / -path "*/openclaw/dist/control-ui/assets/index-*.js" 2>/dev/null | head -1

Windows (PowerShell):

Get-ChildItem -Path C:\ -Recurse -Filter "index-*.js" -ErrorAction SilentlyContinue | Where-Object { $_.FullName -match "openclaw\\dist\\control-ui\\assets" } | Select-Object -First 1 -ExpandProperty FullName

Quick alternative: Run npm root -g to find the global node_modules path, then look inside for openclaw/dist/control-ui/assets/index-*.js.

If you (the AI agent) cannot find or access the file, tell the user:

"I couldn't access the OpenClaw Control UI file. Can you help? Either: 1. Copy the file into my workspace so I can patch it, then copy it back after. Run this in your terminal: cp /path/to/openclaw/dist/control-ui/assets/index-*.js ~/.openclaw/workspace-allgemein/ 2. Or give me write access: sudo chown -R $(whoami) /path/to/openclaw/dist/"

Keep it simple. Don't over-explain.

Step 2: Back up

cp <path-to-index-file> <path-to-index-file>.bak

Step 3: Verify compatibility

Before patching, confirm these patterns exist in the file. If any are missing, the OpenClaw version has changed — do NOT proceed. Instead, inspect the file and adapt, or wait for an updated version of this skill.

PatternWhat it is
function NC(e)Speech recognition start function
sttRecordingRecording state variable
N.micOffMic-off icon reference
jC()Speech recognition availability check

Step 4: Add sttRecordingCont state variable

Find this exact text:

sttRecording:!1,sttInterimText

Replace with:

sttRecording:!1,sttRecordingCont:!1,sttInterimText

This appears exactly once in the file.

Verify: Search for sttRecordingCont:!1 — must appear exactly once.

Step 5: Show interim text for both recording states

Find this exact text:

${X.sttRecording&&X.sttInterimText?i`<div class="agent-chat__stt-interim">${X.sttInterimText}</div>`:h}

Replace with:

${(X.sttRecording||X.sttRecordingCont)&&X.sttInterimText?i`<div class="agent-chat__stt-interim">${X.sttInterimText}</div>`:h}

This appears exactly once in the file.

Verify: Search for (X.sttRecording||X.sttRecordingCont) — must appear in the line containing stt-interim.

Step 6: Replace single mic button with hold-for-continuous button

This is the most complex patch. Replace the ENTIRE mic button block with a single button that supports both short-click (auto-send) and hold-3s (continuous) modes.

6.1 Find the block

Search for the text Stop recording or Voice input in the file. This is inside the mic button template.

The block to replace starts at ${jC()?i and ends at :h} — both on the same line as Stop recording/Voice input, or on the lines immediately surrounding it.

It contains exactly ONE <button> element with:

  • class containing agent-chat__input-btn
  • @click handler referencing X.sttRecording and NC({
  • title with Stop recording / Voice input
  • N.micOff and N.mic for the icon

6.2 Replace the entire block

Select everything from ${jC()?i through :h} inclusive and replace it with:

${jC()?i`
                  <button
                    class="agent-chat__input-btn ${(X.sttRecording||X.sttRecordingCont)?`agent-chat__input-btn--recording`:``}"
                    @mousedown=${(t)=>{if(X.sttRecording||X.sttRecordingCont)return;t.preventDefault();X._holdTimer=setTimeout(()=>{X._holdTimer=null;X._holdTriggered=!0;NC({onTranscript:(t,n)=>{if(n){let n=_(),r=n&&!n.endsWith(` `)?` `:``;e.onDraftChange(n+r+t),X.sttInterimText=``,e.onSend()}else X.sttInterimText=t;g()},onStart:()=>{X.sttRecordingCont=!0;window.__sttCont=MC;g()},onEnd:()=>{if(X.sttRecordingCont){let r=window.__sttCont;if(r){try{r.start()}catch(e){}}X.sttInterimText=``,g()}else{X.sttInterimText=``,g()}},onError:()=>{X.sttRecordingCont=!1,X.sttInterimText=``,g()}})&&(X.sttRecordingCont=!0,g())},3000)}}
                    @mouseup=${(t)=>{if(X._holdTimer){clearTimeout(X._holdTimer);X._holdTimer=null;if(!X._holdTriggered){NC({onTranscript:(t,n)=>{if(n){let n=_(),r=n&&!n.endsWith(` `)?` `:``;e.onDraftChange(n+r+t),X.sttInterimText=``,e.onSend(),PC(),X.sttRecording=!1}else X.sttInterimText=t;g()},onStart:()=>{X.sttRecording=!0;if(MC)MC.continuous=!1;g()},onEnd:()=>{X.sttRecording=!1,X.sttInterimText=``,g();const d=_();if(d&&d.trim())e.onSend()},onError:()=>{X.sttRecording=!1,X.sttInterimText=``,g()}})&&(X.sttRecording=!0,g())}}X._holdTriggered=!1}}
                    @click=${()=>{if(X.sttRecordingCont){PC(),X.sttRecordingCont=!1,window.__sttCont=null,X.sttInterimText=``,g()}}}
                    title=${X.sttRecordingCont?`Stop voice continuous`:(X.sttRecording?`Stop voice auto`:`Voice (hold 3s for continuous)`)}
                    ?disabled=${!e.connected}
                  >
                    ${(X.sttRecording||X.sttRecordingCont)?N.micOff:N.mic}
                  </button>
                `:h}

Do NOT replace individual pieces. Replace the ENTIRE block as a single unit.

Note for the edit tool: Escape < → \<, > → \>, & → \& in the replacement text.

6.3 How it works

mousedown starts a 3-second timer:

  • If mouseup before 3s → timer cancelled, short click → auto-send mode starts (one-shot, continuous=false, auto-send + auto-stop on final transcript or silence)
  • If timer fires after 3s → continuous mode starts (mic stays on, onEnd restarts recognition via window.__sttCont.start(), text auto-sent after each pause)

click (separate from mousedown/mouseup): If already in continuous mode, stops it (PC() + clear window.__sttCont).

Auto-send mode also auto-stops if the user says nothing for ~3 seconds — the browser's SpeechRecognition fires onEnd after silence when continuous=false, which acts as a natural "misclick" safeguard.

6.4 Verify

Search the file for ALL of these strings — every single one must be present:

  • [ ] sttRecordingCont — appears multiple times
  • [ ] _holdTimer — hold timer variable
  • [ ] _holdTriggered — hold flag
  • [ ] Voice (hold 3s for continuous) — idle tooltip
  • [ ] Stop voice auto — auto-send active tooltip
  • [ ] Stop voice continuous — continuous active tooltip
  • [ ] window.__sttCont — continuous restart mechanism
  • [ ] MC.continuous=!1 — one-shot mode flag

If any are missing, restore from backup and retry from step 6.2.

Step 7: Restart and verify

Tell the user:

  1. Restart the gateway (e.g. openclaw gateway restart)
  2. Hard-refresh the browser with Ctrl+Shift+R (normal refresh is NOT sufficient — the browser caches the old JS file)

After the user has done both, verify:

  1. One mic button with tooltip "Voice (hold 3s for continuous)"
  2. Short click → speak → pause → message sent automatically → mic off
  3. Short click → say nothing → mic auto-stops after ~3 seconds
  4. Hold button 3 seconds → continuous mode activates (tooltip changes to "Stop voice continuous")
  5. In continuous mode: speak → pause → message sent → mic stays on → speak again → sent → ...
  6. Click button in continuous mode → mic stops
  7. Live transcript appears above input field while mic is active

Rollback

If anything goes wrong: cp <path-to-index-file>.bak <path-to-index-file> and tell the user to restart the gateway.

Troubleshooting

ProblemLikely CauseFix
No mic button at allBrowser doesn't support Web Speech APIUse Chrome or Edge
Button appears but does nothing on clickStep 6 not applied or partialRestore backup, re-apply Step 6
Short click starts mic but doesn't auto-sende.onSend() missing in @mouseup handlerRestore backup, re-apply Step 6
Short click mic stays on after pausePC() or sttRecording=!1 missingRestore backup, re-apply Step 6
Hold 3s doesn't start continuous mode_holdTimer / _holdTriggered not workingRestore backup, re-apply Step 6
Continuous mode stops after silenceonEnd restart logic not workingCheck window.__sttCont and __sttCont.start() in Step 6 replacement
Live transcript not showingStep 5 not appliedApply Step 5
"Microphone denied" in browserPermissions-Policy blocks micFix microphone=() → microphone=(self) in the HTTP utils file (included in OpenClaw 2026.4.18+)
After OpenClaw updatePatches overwrittenRe-apply ALL steps (filename hash changes!)

平台分布

OpenClaw

75.08%
按下载量换算955

安全审计

VirusTotal

可疑

ClawScan

通过

Static analysis

通过

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills