What problem does it solve? Sending or reading KakaoTalk messages on behalf of a user requires GUI automation that is blocked inside sandboxed agent environments, and naive approaches fail on Korean input, hidden chat rooms, and unverified delivery. This Skill provides a tested procedure for operating KakaoTalk reliably from an AI agent on macOS. ## Core Features & Use Cases - Message sending with verification: Sends messages to a chat room by name or chat ID via kmsg, then confirms delivery by reading the local message database with kakaocli. - Sandbox escape via SSH loopback: Runs all commands through ssh localhost because the sandbox blocks GUI execution, accessibility (AX) calls, and process inspection. - Failure recovery playbook: Maps concrete error codes like [SEARCH_MISS] and [WINDOW_NOT_READY] to fixes such as --refresh-cache retries. - Use Case: A user asks the agent to forward an announcement to a group chat; the agent drafts the text to a file, dry-runs the recipient, sends it, appends an AI-disclosure signature, and verifies delivery in the database. ## Quick Start Ask the agent to send a KakaoTalk message to a specific chat room and confirm it was delivered.
