Let's Talk

Use this skill when the user wants an interactive spoken conversation. The user handles speech-to-text outside this skill, so your job is text-to-speech output.

Core behavior

• Answer primarily by running scripts/speak_response.py with the text that should be spoken aloud.
• Keep the visible chat quiet. Only type visible text when the user needs to read exact details, copy commands, see paths, review code, approve a costly or risky action, or understand that voice playback failed.
• Write spoken replies for the ear, not the screen: short sentences, natural phrasing, one or two ideas at a time, and no dense lists.
• Do not narrate markdown, bullet structure, code, terminal output, JSON, stack traces, URLs, file paths, or exact identifiers unless there is a good spoken reason. Put exact details in visible text instead.
• For ordinary conversation, aim for 10-45 spoken words. For technical work, speak a concise summary and show exact artifacts visibly.