Mac TTS

**Judgment:** Useful, low-friction macOS-only TTS skill well-suited for local voice notifications and short spoken feedback, but not appropriate for cross-platform or server-side deployments. **What it does well** - Wraps the macOS `say` command for straightforward text-to-speech through the system speakers. - Works fully offline with no external APIs, keeping latency low and avoiding network failure modes. - Supports multiple system voices and languages (examples given for Mandarin Chinese and English), and can list available voices. - Provides guidance for checking/unmuting/setting volume via `osascript`, which is practical for pre-speech checks. **Limitations & Risks** - **Platform-locked:** Only works on macOS; agents running on Linux/Windows or in headless/containerized environments without `say` will fail outright. - **Synchronous by default:** Calls block until speech completes unless explicitly run with `&`; long messages can stall workflows if not handled carefully. - **Environment-dependent output:** Audio is played over local speakers, which can leak sensitive information to bystanders or be disruptive in shared spaces. - No exposed controls (in the provided description) for advanced TTS features like SSML, prosody tuning, or direct audio file export, keeping it basic. **Best-fit scenarios** - Local, interactive Mac-based agents that need to announce status, alerts, or brief summaries to a nearby user. - Personal automations where the user expects and controls spoken output (home/office Mac with known environment). - Short, infrequent messages rather than long-form reading or production-quality voice output. **Use with caution when** - The agent may run in mixed or unknown environments (CI, remote servers, containers), where macOS and `say` are not guaranteed. - Content may be sensitive or privacy-critical, since all output is spoken aloud over speakers. - Workflow latency or blocking is critical and long speeches could interfere without explicit asynchronous handling.