speak-hello-world — Claude Code skill: language tutoring with pronunciation feedback
speak-hello-world — a skill for Claude Code, Codex, Cursor, and other platforms that sets up a language tutoring session via the Speak API with pronunciation scoring and real-time feedback.
MIT 1 installs ★ 2,935 updated 2026-10-29
npx skills add jeremylongshore/tons-of-skills-marketplace --skill speak-hello-world -a claude-codeWhat speak-hello-world does
The skill spins up a conversation session with an AI tutor, sends student responses as text or audio, rates pronunciation at the phoneme and word level, flags grammar mistakes, and tracks new vocabulary. The code includes a TypeScript client that initializes the session, processes replies, and ends with learning metrics.
You need valid credentials from speak-install-auth already in place — the API key and App ID go into SPEAK_API_KEY and SPEAK_APP_ID environment variables. Supports Spanish, Korean, Japanese, French, and German. Pronunciation scoring expects WAV files at 16 kHz mono; other codecs or sample rates will trigger an error.
What it helps with
- set up your first language tutoring session with Speak
- score pronunciation with phoneme-level breakdown
- get grammar corrections and vocabulary summary after a lesson
- verify your Speak API setup before building the full app
- practice dialogue in multiple languages
Example prompts for the agent
- Create a beginner Spanish lesson on greetings and score my pronunciation
- Run speak-hello-world to test whether my Speak API credentials work
- Test the flow with text responses only, no audio or microphone
How to install speak-hello-world
1Skills CLI (Vercel)
npx skills add jeremylongshore/tons-of-skills-marketplace --skill speak-hello-world -a claude-code2Manually: put the files in place
~/.claude/skills/speak-hello-world/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/speak-hello-world/
1Skills CLI (Vercel)
npx skills add jeremylongshore/tons-of-skills-marketplace --skill speak-hello-world -a codex2Manually: put the files in place
.agents/skills/speak-hello-world/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add jeremylongshore/tons-of-skills-marketplace --skill speak-hello-world -a cursor2Manually: put the files in place
.cursor/skills/speak-hello-world/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install jeremylongshore/tons-of-skills-marketplace skills/.curated/speak-hello-world2Manually: put the files in place
.github/skills/speak-hello-world/SKILL.md1Skills CLI (Vercel)
npx skills add jeremylongshore/tons-of-skills-marketplace --skill speak-hello-world -a gemini-cli2Platform CLI
gemini skills install https://github.com/jeremylongshore/tons-of-skills-marketplace --path skills/.curated/speak-hello-worldManually: put the files in place
.devin/skills/speak-hello-world/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add jeremylongshore/tons-of-skills-marketplace --skill speak-hello-world -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about speak-hello-world
How do I install speak-hello-world in Claude Code?
Run npx skills add jeremylongshore/tons-of-skills-marketplace --skill speak-hello-world -a claude-code, then open ~/.claude/skills/speak-hello-world/SKILL.md. Commands differ by platform: for Codex use npx skills add jeremylongshore/tons-of-skills-marketplace --skill speak-hello-world -a codex, for Cursor use npx skills add jeremylongshore/tons-of-skills-marketplace --skill speak-hello-world -a cursor, for GitHub Copilot use gh skill install jeremylongshore/tons-of-skills-marketplace speak-hello-world.
What does speak-hello-world do?
It launches a tutoring session with an AI tutor, processes student responses, rates pronunciation at phoneme and word level, reports grammar errors and new words, and closes the session with learning stats.
What errors might I hit?
If audio is in the wrong format, convert to WAV 16 kHz mono. When a session times out, start a new one. If a language is not supported, pick from the list: es, ko, ja, fr, de. Low pronunciation scores usually mean background noise; record in a quiet space.
Source text
SKILL.md · MIT
--- name: speak-hello-world description: 'Create your first Speak AI tutoring session with pronunciation feedback. Use when starting a new Speak integration, testing your setup, or learning basic language learning API patterns. Trigger with phrases like "speak hello world", "speak example", "speak quick start", "first speak lesson". ' allowed-tools: Read, Write, Edit, Bash(npm:*), Bash(node:*) version: 1.14.0 license: MIT author: Jeremy Longshore <[email protected]> tags: - saas - speak - api - testing compatibility: Designed for Claude Code --- # Speak Hello World ## Overview Create your first AI tutoring session with Speak. Demonstrates conversation practice, pronunciation assessment, and real-time feedback using GPT-4o-powered tutoring. ## Prerequisites - Completed `speak-install-auth` setup - Valid API credentials configured - Microphone access (optional for testing) ## Instructions ### Step 1: Start a Conversation Session ```typescript import { SpeakClient } from '@speak/language-sdk'; const client = new SpeakClient({ apiKey: process.env.SPEAK_API_KEY!, appId: process.env.SPEAK_APP_ID!, language: 'es', }); // Start a beginner Spanish lesson const session = await client.startConversation({ scenario: 'greetings', language: 'es', level: 'beginner', nativeLanguage: 'en', }); console.log('Session ID:', session.id); console.log('AI Tutor:', session.firstPrompt.text); // Output: "Hola! Bienvenido a tu leccion de espanol. Como te llamas?" console.log('Audio URL:', session.firstPrompt.audioUrl); ``` ### Step 2: Send a Student Response ```typescript // Submit text response (or audio file for pronunciation scoring) const turn = await client.sendTurn(session.id, { text: 'Hola, me llamo Juan. Mucho gusto.', // Or: audioPath: './recordings/response.wav' }); console.log('Tutor response:', turn.tutorText); console.log('Pronunciation score:', turn.pronunciationScore); // 0-100 console.log('Grammar corrections:', turn.corrections); // Output: [{original: "me llamo", suggestion: null, correct: true}] console.log('Vocabulary notes:', turn.vocabularyNotes); ``` ### Step 3: Pronunciation Assessment ```typescript // Assess pronunciation of a specific phrase const assessment = await client.assessPronunciation({ audioPath: './recordings/hola-como-estas.wav', targetText: 'Hola, como estas?', language: 'es', detailLevel: 'phoneme', // 'word' or 'phoneme' }); console.log(`Overall score: ${assessment.score}/100`); for (const word of assessment.words) { console.log(` "${word.text}": ${word.score}/100`); if (word.phonemes) { for (const p of word.phonemes.filter(p => p.score < 70)) { console.log(` Weak phoneme: ${p.symbol} (${p.score}) - ${p.suggestion}`); } } } ``` ### Step 4: End Session and Review ```typescript const summary = await client.endSession(session.id); console.log('Session Summary:'); console.log(` Duration: ${summary.durationMinutes} min`); console.log(` Turns: ${summary.totalTurns}`); console.log(` Pronunciation: ${summary.avgPronunciationScore}/100`); console.log(` Grammar: ${summary.grammarAccuracy}%`); console.log(` New vocabulary: ${summary.newWords.join(', ')}`); ``` ## Output - Working conversation session with AI tutor - Pronunciation assessment with phoneme-level feedback - Session summary with learning metrics - Console output showing scores and corrections ## Error Handling | Error | Cause | Solution | |-------|-------|----------| | Session timeout | Exceeded max duration | Start a new session | | Audio format invalid | Wrong codec or sample rate | Convert to WAV 16kHz mono | | Language not supported | Invalid language code | Use supported codes (es, ko, ja, fr, de) | | Low pronunciation score | Background noise | Record in a quiet environment | | Rate limit exceeded | Too many requests | Wait and retry with backoff | ## Resources - [Speak Website](https://speak.com) - [Speak GPT-4 Blog](https://speak.com/blog/speak-gpt-4) - [OpenAI Realtime API](https://platform.openai.com/docs/guides/realtime) ## Next Steps Proceed to `speak-local-dev-loop` for development workflow setup. ## Examples **Text-only test**: Skip audio and use text responses to test the conversation flow before integrating microphone input. **Multi-language**: Start sessions in different languages by changing the `language` parameter to `ko` (Korean), `ja` (Japanese), or `fr` (French).
© Jeremy Longshore · MIT · jeremylongshore/tons-of-skills-marketplace