phone-call

Makes outbound SIP phone calls and conducts AI voice conversations returning structured data.

14|Updated Mar 2, 2026
One-click install
npx skills add https://github.com/intendant-dev/Intendant --skill phone-call-intendant-dev
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: phone-call
Source: https://github.com/intendant-dev/Intendant/tree/main/skills/phone-call
Command: npx skills add https://github.com/intendant-dev/Intendant --skill phone-call-intendant-dev

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Automating real phone calls is hard: you need SIP telephony, audio routing, and a conversational model working together. This Skill places an outbound SIP call through pjsua, routes audio via the Vortex Audio virtual device, and lets an AI model conduct the conversation, returning typed structured data extracted from the call. ## Core Features & Use Cases - Outbound SIP Calling: Starts a pjsua session with SIP credentials and routes call audio through the Vortex Audio HAL device on macOS. - AI Voice Conversation: Uses spawn_live_audio with a playbook so the model speaks and listens during the live call. - Structured Result Extraction: A mandatory response_schema validates the conversation output into typed fields (string, integer, boolean, array) with statuses like Completed, TimedOut, or SchemaError. - Use Case: Ask the agent to call a restaurant and book a table; it returns structured fields such as guest name, party size, reservation time, and confirmation status. ## Quick Start Ask the agent to call a phone number and collect specific information, for example: call this restaurant and confirm a reservation for four people at 7 PM tonight.

Frequently Asked Questions about phone-call

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make an automated phone call with an AI voice agent?▼

Start pjsua with your SIP credentials and the Vortex Audio device index, then immediately call spawn_live_audio with a playbook and a response_schema. The model conducts the conversation and returns structured data matching your schema.

How do I extract structured data from a phone call conversation?▼

Define a response_schema with typed fields (string, integer, boolean, array) for every piece of data you want. The model's spoken output is validated against this schema and returned as JSON when the call completes.

Does this phone call skill work on Linux or Windows?▼

No, it is macOS only. It requires the Vortex Audio HAL plugin, pjsua at ~/bin/pjsua, and a GUI session with TCC microphone permission, plus a reachable Intendant daemon.

Why was my phone call rejected with a parse error?▼

The call is rejected when spawn_live_audio is invoked without a response_schema. The schema is mandatory; every field you want extracted from the conversation must be declared with a name, type, and description.

What happens if the phone call times out or disconnects?▼

spawn_live_audio returns a LiveAudioResult with a status field: TimedOut when the call exceeds timeout_secs (default 300), Disconnected when the live connection ends early, or Failed for setup and audio processing errors.