Design: add macOS system and microphone audio capture

Motivation

Native SDK exposes audio playback but not audio input. Meeting recorders and similar apps currently need an app-specific native helper for system and microphone capture.

Proposed scope

• macOS 15+ AppKit system-host implementation.
• Capture system audio, a selected/default microphone, or both into one signed 16-bit PCM WAV.
• Enumerate connected microphones and emit a devices-changed invalidation.
• Explicit permission status/request effects; capture start never prompts.
• Zig and TypeScript command/event parity with replay-safe fixed records.
• Manifest privacy strings and generated macOS Info.plist entries.
• Honest feature discovery and unsupported results on older macOS and other hosts.

Proposed API

Features: system_audio_capture, microphone_capture, and microphone_device_enumeration.

Zig effects: startAudioCapture, stopAudioCapture, listMicrophoneDevices, and audioCaptureAccess.

TypeScript commands: audioCaptureStart, audioCaptureStop, microphoneDevices, audioCaptureAccess, plus a microphoneDevicesChanged subscription.

Native implementation

Use ScreenCaptureKit for system-only and combined capture, and AVFoundation for microphone-only capture. Timestamp-align and resample buffers, mix combined sources with headroom, write through a sibling temporary file, and atomically publish the finalized WAV. Allow one active capture and preserve a finalized partial file after asynchronous failure.

Packaging

Add a logical system_audio permission plus microphone_usage and system_audio_usage privacy strings. Generate the relevant microphone, ScreenCaptureKit, and system-audio purpose keys. App Sandbox remains opt-in; document the audio-input entitlement.

Out of scope

Raw PCM delivery, meters, video capture, per-app audio filtering, source gains, and app-specific meeting UI.

Start by locating the Zig effects and TypeScript command/event entry points named in the proposal, then trace the macOS native integration for ScreenCaptureKit and AVFoundation. Define how capture, device enumeration, permissions, replay-safe records, packaging privacy keys, and unsupported hosts should behave; done means the proposed API and native behavior are implemented end to end.