ElevenLabs ships its whole audio API as agent skills

What it does This repo packages ElevenLabs’ developer products as skills following the Agent Skills specification, so any compatible AI coding assistant can invoke them. Coverage spans the full audio surface: text-to-speech, speech-to-text, real-time voice conversations, conversational agents, sound effects, music generation, voice changing, voice isolation, dubbing, plus onboarding and API-key setup skills. Most skills ship examples for Python, JavaScript/TypeScript, and the ElevenLabs CLI.

The interesting bit It’s not just prompt files — the repo includes an evals/ harness with trigger tests (does the skill fire for the right queries?) and functional tests (does it produce correct output?), run in isolated cursor-agent workspaces. That’s a level of testing discipline most skill collections skip entirely.

Key highlights

• Twelve skills covering TTS, STT, realtime voice, agents, music, SFX, dubbing, and voice transformation
• Works with any Agent Skills–spec-compatible assistant, not just one vendor’s tooling
• Trigger and functional evals for every skill, with machine-readable results in results.json
• onboarding and setup-api-key skills handle the API-key plumbing, including admin-issued keys that break
• Explicit warning against the outdated npm install elevenlabs v1 package in favor of @elevenlabs/elevenlabs-js

Caveats

• Everything requires an ElevenLabs API key and hits their cloud — nothing here runs offline
• Running the evals yourself requires the Cursor Agent CLI and Cursor authentication, which quietly couples the test suite to one vendor

Verdict Worth adding if you’re building anything voice-adjacent with an agent-capable assistant — it turns a sprawling audio API into one-line invocations. If you’re not already in the ElevenLabs ecosystem, this is a sales funnel with documentation, not a toolbox you need.