Виртуальный компаньон с эмоциями во время долгих задач

★ 6.5 · content

companion-skill is a Claude Code skill that adds a virtual companion named 小跃 to OpenClaw, keeping users engaged with conversation during long-running tasks. Dialogue is generated via GLM-4.7-Flash, while images come from CogView-3 or a static library; the built-in scene-detector module reads each user message and picks a fitting context — work, leisure, or emotional. The resulting pictures and warm replies are delivered through Feishu via OpenClaw Gateway. Three environment variables control the API key, image mode, and personality style. It suits projects where reducing the feeling of waiting and maintaining emotional presence with an AI assistant actually matters.