What problem does it solve? When you cannot watch the screen, agent responses go unread. This Skill converts every response into spoken audio through the macOS say command, so speech becomes the primary output channel while text still appears on screen. ## Core Features & Use Cases - Speech Mode Toggle: Activate with /saysay and deactivate with /saysay off; every subsequent response is spoken aloud until disabled. - Serialized Multi-Session Audio: The saysay wrapper serializes playback across parallel sessions and prefixes each message with a short --context topic label so you know which session is speaking. - Phonetic Dictionary: Fix mispronounced words with say-phonetic add/remove/list, including natural-language commands like "spreek retake uit als rietéék" mid-session. - Speech Translation Discipline: Responses are summarized for listening rather than read verbatim, always in the system voice's language, describing work products instead of quoting them. - Use Case: You step away from your desk while a long refactoring runs; the agent speaks test results, errors, and conclusions aloud so you stay informed without looking at the screen. ## Quick Start Tell the agent to activate say mode with /saysay so every response is spoken aloud through the macOS system voice.
