New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Voice Command Or Chatter

by mrmps in mrmps/classifier-dev

Decide whether a speech transcript line is aimed at the assistant or at the other people in the room.

MITAuto-check passedAI & LLM Engineering

Install Voice Command Or Chatter

skills CLI
$ npx skills add mrmps/classifier-dev --skill voice-command-or-chatter -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install mrmps/classifier-dev voice-command-or-chatter --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/mrmps/classifier-dev.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/voice-command-or-chatter .claude/skills/voice-command-or-chatter && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
voice-command-or-chatter
GitHub stars
424
Token cost
~1.4k tokens
SKILL.md length
708 words
Files
1
Skills in repo
21
Repo updated
First seen
Licence
MIT

At a glance

Decide whether a speech transcript line is aimed at the assistant or at the other people in the room.

  • Works in 4 steps: One line, to check the wiring → A real run from the shell → The gate → …
  • Any ASR to an agent
  • SKILL.md covers When not to use it, 1. One line, to check the wiring, 2. A real run from the shell and 3. The gate, plus 3 more sections
  • Calls curl and npm; reaches classifier.dev

What it does

Voice Command Or Chatter is an agent skill from mrmps/classifier-dev. Decide whether a speech transcript line is aimed at the assistant or at the other people in the room. Classifies each utterance as command, question, chatter or partial with a calibrated confidence, so an always-on microphone wakes on the two that are addressed to it and stays quiet through the rest. Use when wiring Whisper or any ASR to an agent, building a voice assistant, meeting bot or car interface, or on "it answers things nobody asked it", "false wakes", "it fires halfway through a sentence".

Its SKILL.md is about 1.4k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in AI & LLM Engineering, covering Speech recognition and synthesis. The repository describes itself as: Zero-shot text classification over plain HTTP — no API key, no account. One Cloudflare Worker, a CLI, and an MCP server. https://classifier.dev. The licence is MIT.

When your agent uses it

  • Any ASR to an agent
  • Building a voice assistant
  • On it answers things nobody asked it
  • It fires halfway through a sentence

Example prompts

  • “it answers things nobody asked it”
  • “false wakes”
  • “it fires halfway through a sentence”
  • “/voice-command-or-chatter”

Requirements

  • Node.js

Workflow steps

4 steps, taken from the step headings in SKILL.md.

  1. One line, to check the wiring
  2. A real run from the shell
  3. The gate
  4. The sliding window over partials

What it can do on your machine

Read from SKILL.md and the folder at commit 629df75. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • curl
    • npm

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    Hosts in commands or code, which the agent is likely to contact:

    • classifier.dev

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Voice Command Or Chatter loads about 1.4k tokens when it runs. Until then it costs about 132 tokens; SKILL.md has 708 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~132
When it runs · the whole SKILL.md, loaded when a task matches
~1.4k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from mrmps/classifier-dev at commit 629df75, republished under its MIT licence (© mrmps). 708 words, ~1,411 tokens.

Download SKILL.mdSave it as .claude/skills/voice-command-or-chatter/SKILL.md (or your agent's skills folder).
name
voice-command-or-chatter
description
Decide whether a speech transcript line is aimed at the assistant or at the other people in the room. Classifies each utterance as command, question, chatter or partial with a calibrated confidence, so an always-on microphone wakes on the two that are addressed to it and stays quiet through the rest. Use when wiring Whisper or any ASR to an agent, building a voice assistant, meeting bot or car interface, or on "it answers things nobody asked it", "false wakes", "it fires halfway through a sentence".
license
MIT

Wake only on what was said to you

Continuous ASR gives you every word in the room, including the half of it that was never meant for the assistant. A wake word solves that by making people talk like a remote control. Classifying the utterance instead lets them talk normally, and gives you a number to gate on.

Four labels carry it: command, question, chatter, partial.

When not to use it

  • With a wake word already in the pipeline. The wake word did the job.
  • For what the command means. This says the line was addressed to you; your own model turns it into an intent and slots.
  • On audio. It reads text, so ASR comes first, and its errors are yours.
  • Where a false wake is expensive — a car control, a payment. Raise the gate and confirm out loud.

1. One line, to check the wiring

curl -s "https://classifier.dev/command,question,chatter,partial/turn+off+the+kitchen+lights"
command

2. A real run from the shell

npm i -g classifier-dev@0.1.3

transcript.txt, one utterance per line, straight out of the ASR:

classify command,question,chatter,partial \
  -i "Each line is one utterance from a live transcript of a room with a voice assistant in it. A command asks the assistant to do something, a question asks it for information, chatter is speech between people, and partial is a fragment cut off mid-utterance." \
  < transcript.txt

Real output, label<TAB>confidence<TAB>text:

command	1.00	turn off the kitchen lights
question	1.00	what time does the pharmacy close
chatter	1.00	so anyway i told him it was fine and he just laughed
partial	1.00	i think the
chatter	0.55	no no i meant the other one, the blue one
command	1.00	play something quiet
chatter	0.98	yeah okay so we should probably leave around six
command	0.83	can you add milk to the shopping

That instructions sentence is doing real work: it is what makes "so anyway I told him" chatter rather than a command to tell someone something. Write it once, from your own product's point of view, and keep it fixed.

3. The gate

Wake on command or question at or above 0.85, and only then.

  • 0.85 and up — wake. Three of the eight lines above. Three more are chatter or partial at 0.98 or better and never wake it at any score.
  • 0.5 to 0.85 — do not wake; keep the line. Line 8 lands at 0.83 and line 5 at 0.55. Hold them in the window (below) and re-score when the next words arrive; a real request is almost always completed.
  • Below 0.5 — drop it. This is chatter you were never part of.
  • chatter or partial — stay asleep, whatever the confidence.

0.85 rather than the usual 0.9 because a missed wake costs a repeat and a false wake costs trust in the microphone; move it up, not down, if the room is noisy.

Do not put tier: "smart" in the wake path. It escalates answers under 0.7 to a reasoning model, and on line 5 it did: escalated: true, still chatter, 0.52, usage.ms 1,713 against about 150 ms on fast. A second and a half is the whole latency budget of a voice turn. Use smart offline, on yesterday's transcripts, to find where the gate was wrong.

Show full SKILL.md (305 more words)Show less

4. The sliding window over partials

Streaming ASR emits a growing prefix several times a second. Classifying every emission wakes the assistant mid-sentence. Instead:

  • Keep a buffer of the current utterance. Classify only when the ASR reports a segment boundary, or after about 600 ms of silence.
  • Re-score the buffer on each new boundary, not each token.
  • partial above 0.85 means the sentence is not finished: keep buffering, do not wake, do not clear.
  • Clear the buffer on a wake, or after two seconds of silence.

The same four labels over one utterance as it grows:

partial	0.57	set a
command	0.67	set a timer
command	0.55	set a timer for twelve
command	1.00	set a timer for twelve minutes

Only the complete sentence clears 0.85. That is the behaviour you want, and it is why the gate goes on the buffer rather than on each emission.

Pitfalls

  • A line near the gate moves between runs. Line 8 came back 0.83, 0.87 and 0.80 on three runs of the same file. Re-score the finished utterance; do not wake on the first crossing of a growing prefix.
  • Every call returns one of your four labels. Music, a television and a one-sided phone call all land somewhere. If those are common in the room, add a fifth label such as background audio or someone else's phone call.
  • Scores do not validate ASR input. Include labels for punctuation, noise markers and other expected artifacts. Treat any unavailable score as "do not wake", never as zero.
  • Batch the offline pass. Up to 1,000 utterances per call and 3,000 classifications a minute per IP; a whole day of transcript is a few calls.

What done looks like

Every utterance has a label and a confidence in the log. The assistant wakes only on command or question at 0.85 or higher, a growing partial never wakes it, and the 0.5 to 0.85 lines are kept for a weekly look at where the gate was wrong.

© mrmps, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in skills/voice-command-or-chatter of mrmps/classifier-dev.

Open the folder on GitHubat commit 629df75

Compare with similar skills

Voice Command Or Chatter next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Voice Command Or Chatter compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Voice Command Or Chatter this skillmrmps/classifier-dev424—~1.4kAutomated safety check: PassMIT
TriageTalAter/annyang6.8k1 repos~810Automated safety check: NotesMIT
Yichen Asrmcncarl/yichen-skills4.3k—~780Automated safety check: PassCustom licence
Dingtalk MinutesDingTalk-Real-AI/dingtalk-workspace-cli3.2k—~2.3kAutomated safety check: PassApache-2.0
Youtube FetcherJimmySadek/youtube-fetcher-to-markdown485—~3.1kAutomated safety check: PassMIT
Yichen Web Researchmcncarl/yichen-skills4.3k—~1.9kAutomated safety check: PassCustom licence

Similar skills

  • Triage

    TalAter/annyang

    Triage and close GitHub issues on TalAter/annyang. An agent skill from TalAter/annyang.

    6.8k GitHub starsUsed in 1 repo~810 tokens
    AI & LLM EngineeringAuto-check: notes

  • Yichen Asr

    mcncarl/yichen-skills

    逸尘自用的统一音视频转写入口，在 StepFun Step ASR 与火山引擎豆包 ASR 之间按输出需求、安全边界和可用状态路由。用于本地音频或视频的纯文本转写、时间戳、SRT 字幕、口播粗剪，以及转写前体检；用户明确指定服务商时不得静默切换。Use when a local audio or video file needs transcription and the correct…

    4.3k GitHub stars~780 tokensUpdated 3 days ago
    AI & LLM EngineeringAuto-check passed

  • Dingtalk Minutes

    DingTalk-Real-AI/dingtalk-workspace-cli

    钉钉 AI 听记。Use when 查询或修改听记摘要、完整逐字稿、关键词、标签、行动项、录音、上传、思维导图、发言人洞察、ASR 热词/识别词配置或分享权限。写文档走 dingtalk-doc；建待办走 dingtalk-todo；日程走 dingtalk-calendar。命令前缀：dws minutes。

    3.2k GitHub stars~2.3k tokensUpdated 2 days ago
    AI & LLM EngineeringAuto-check passed

  • Youtube Fetcher

    JimmySadek/youtube-fetcher-to-markdown

    Retrieve transcripts from YouTube, Instagram, TikTok, X, Vimeo and other video sites, summarize or analyze what was said (and shown on screen), or save an Obsidian-ready Markdown knowledge-base note…

    485 GitHub stars~3.1k tokensUpdated today
    AI & LLM EngineeringAuto-check passed

  • Yichen Web Research

    mcncarl/yichen-skills

    逸尘自用的互联网研究总入口。用于跨平台且跨阶段、用户尚未确定工具，或明确要求对公司、产品、人物、技术、行业和领域做横纵分析、发展史加现状对比或有来源约束的系统深度研究；先生成有截止日期和证据闸门的计划，再把搜索发现、候选核验、有限归档、按需转写和证据综合路由到…

    4.3k GitHub stars~1.9k tokensUpdated 3 days ago
    AI & LLM EngineeringAuto-check passed

  • Volcengine Asr

    ysyecust/lecture-to-notes

    Transcribe local audio or video with Volcengine Doubao file ASR, including BigASR 1.0 Turbo direct upload and asynchronous 1.0 standard, 1.0 idle, or 2.0 standard jobs through TOS.

    270 GitHub stars~783 tokensUpdated 4 days ago
    AI & LLM EngineeringAuto-check passed

More from mrmps/classifier-dev

All 21 skills in this repo

  • Bulk Classify

    mrmps/classifier-dev

    Sort many texts into your own categories without reading them, using a keyless HTTP API that returns a calibrated confidence per answer.

    424 GitHub stars~3.1k tokensUpdated today
    Auto-check passed

  • Computer Use Action Picker

    mrmps/classifier-dev

    Pick a browser or desktop agent's next action by choosing among the actions actually on screen instead of inventing one.

    424 GitHub stars~1.5k tokensUpdated today
    Auto-check passed

  • Content Moderation Gate

    mrmps/classifier-dev

    Check user-generated text against a written policy before it is published.

    424 GitHub stars~1.5k tokensUpdated today
    Auto-check passed

  • Context Compaction Keep Or Drop

    mrmps/classifier-dev

    Label each context chunk keep, drop or replace-with-a-pointer and pass the survivors through byte for byte instead of summarising, with key-shaped chunks decided locally and never sent, and a…

    424 GitHub stars~1.5k tokensUpdated today
    Auto-check passed

  • Document Intake Routing

    mrmps/classifier-dev

    Label each page of an intake packet with a document type and a page role before extraction runs, so only confident pages reach an extractor and the rest reach a person.

    424 GitHub stars~1.5k tokensUpdated today
    Auto-check passed

  • Headline Filter Map Reduce

    mrmps/classifier-dev

    Filter hundreds or thousands of headlines, search results or feed items against a written brief before opening any of them, using a two-stage cascade that spends a fast model on everything and a…

    424 GitHub stars~1.5k tokensUpdated today
    Auto-check passed

Categories

Questions about Voice Command Or Chatter

What does Voice Command Or Chatter do?

Decide whether a speech transcript line is aimed at the assistant or at the other people in the room. Voice Command Or Chatter is an agent skill from mrmps/classifier-dev. Decide whether a speech transcript line is aimed at the assistant or at the other people in the room.

When should I use Voice Command Or Chatter?

Voice Command Or Chatter fits situations like: any ASR to an agent; building a voice assistant; on it answers things nobody asked it; it fires halfway through a sentence.

How do I install Voice Command Or Chatter in Claude Code?

Run `npx skills add mrmps/classifier-dev --skill voice-command-or-chatter -a claude-code`. Or copy the skill folder (skills/voice-command-or-chatter in mrmps/classifier-dev) into .claude/skills/voice-command-or-chatter in your project. Claude Code loads it when a task matches its description.

How do I install Voice Command Or Chatter in Codex?

Run `npx skills add mrmps/classifier-dev --skill voice-command-or-chatter -a codex`. Or copy the skill folder (skills/voice-command-or-chatter in mrmps/classifier-dev) into .agents/skills/voice-command-or-chatter in your project. Codex loads it when a task matches its description.

Can I use Voice Command Or Chatter in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add mrmps/classifier-dev --skill voice-command-or-chatter -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/voice-command-or-chatter, .gemini/skills/voice-command-or-chatter, .github/skills/voice-command-or-chatter and .opencode/skills/voice-command-or-chatter in your project.

What does Voice Command Or Chatter need to run?

Going by SKILL.md and its folder, Voice Command Or Chatter needs the command-line tools its instructions call (curl and npm). Our summary lists: Node.js.

Does Voice Command Or Chatter access the network?

SKILL.md names 1 domain. In commands or code: classifier.dev; the agent is likely to contact it when it follows the instructions. This is read from the text; nothing was executed.

Is Voice Command Or Chatter safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Voice Command Or Chatter use?

Voice Command Or Chatter is published under the MIT licence (declared in SKILL.md). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Voice Command Or Chatter use?

About 1.4k tokens (SKILL.md is roughly 5.6k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Voice Command Or Chatter?

Skills that share tags, products or a category with Voice Command Or Chatter: Triage (TalAter/annyang, 6.8k stars), Yichen Asr (mcncarl/yichen-skills, 4.3k stars), Dingtalk Minutes (DingTalk-Real-AI/dingtalk-workspace-cli, 3.2k stars) and Youtube Fetcher (JimmySadek/youtube-fetcher-to-markdown, 485 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Voice Command Or Chatter?

mrmps (a GitHub user) maintains it in mrmps/classifier-dev, which has 424 GitHub stars. The repository holds 21 skills in this directory. The repository was last updated on October 7, 2026.

Source: mrmps/classifier-dev on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.