agentleFS
IgorGanapolskyIgorGanapolsky/mac-yolo-safeguards/.agents/skills/elevenlabs-voice-agents/SKILL.md

elevenlabs-voice-agents

Skill3 starsChanged 4 days agoMITView on GitHub

ElevenLabs Hosted MCP Voice Agents, Conversational AI Engine, Multi-LLM Cost Estimator & ThumbGate Pre-Action Interdiction Suite for all coding agents on Igor's Mac.

The file

Files
SKILL.md

ElevenLabs Conversational Voice Agents & ThumbGate Governance Suite

This skill implements the ElevenLabs Hosted MCP Connector pattern for voice agents (ElevenAgents), enabling AI coding assistants to configure, simulate, version, and govern production voice agents as code while enforcing ThumbGate pre-action safety guardrails.

Capabilities

  1. Voice-Agent-as-Code (GitOps):
    • Export and version voice agent configurations (system prompts, voice IDs, latency profiles, initial greetings, language targets) as declarative YAML/JSON.
    • Synchronize configurations between git repositories and ElevenLabs workspaces via automated pipelines.
  2. Pre-Deployment Multi-LLM Cost Estimator:
    • Compares expected inference cost per conversation minute across foundation models:
      • Gemini 2.5 Flash (Ultra-low latency conversational tier)
      • Qwen 3.8 / 2.5 (High ROI token plan tier)
      • GLM-5.3 (Zero marginal cost coding plan)
      • GPT-4o (Enterprise flagship tier)
      • Claude 3.5 Sonnet (Reasoning tier)
    • Evaluates token budgets, average turn duration, and tool call overhead before promoting changes.
  3. Conversational Simulation Test Framework:
    • Simulates multi-turn voice dialogs against mock caller intents before production deployment.
    • Asserts tool invocation accuracy, system prompt adherence, and graceful fallback responses.
  4. ThumbGate & Leash Pre-Action Interdiction:
    • Destructive Action Blocking: Intercepts delete_agent, modify_production_prompt, or swap_llm_backend tool calls via ThumbGate PreToolUse hooks.
    • Requires phone Leash operator approval (or prior simulated test pass proof) before executing changes on live customer phone trees.

Global CLI Commands

# Run full voice agent suite audit & status
eleven-voice audit

# Estimate conversation costs across LLM backends (100 calls, 3 mins avg)
eleven-voice cost-estimate --calls 100 --minutes 3

# TNS question: cost with Gemini Flash instead of GPT-4o
eleven-voice compare-models --baseline gpt-4o --candidate gemini-2.5-flash

# Run conversational simulation tests on an agent config
eleven-voice test-agent --config config/voice-agents/hermes-receptionist.v1.json

# Validate ThumbGate pre-action safety gate on a proposed tool call
eleven-voice gate-check --action delete_agent --agent-id agent_prod_01

# Atomic promote (sim → cost → gate). Chat confirm alone never deploys.
eleven-voice promote --config config/voice-agents/hermes-receptionist.v1.json

Configuration Schema

Voice agent configurations follow the standardized declarative format:

{
  "agent_id": "hermes_voice_receptionist_v1",
  "name": "Hermes Mobile Voice Receptionist",
  "conversation_config": {
    "agent": {
      "prompt": {
        "prompt": "You are Hermes, a friendly and ultra-fast voice assistant...",
        "llm": "gemini-2.5-flash",
        "temperature": 0.3
      },
      "first_message": "Hey there! Hermes here. How can I help you today?",
      "language": "en"
    },
    "tts": {
      "voice_id": "21m00Tcm4TlvDq8ikWAM",
      "model_id": "eleven_turbo_v2_5",
      "stability": 0.5,
      "similarity_boost": 0.8
    }
  }
}

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.