Zuma AI Social Media Automation Assistant

Problem Addressed

Zuma AI Social Media Automation Assistant targets repetitive desktop automation for self-media workflows: collecting a specified X/Twitter user, scheduling Xiaohongshu posts, checking execution logs, and capturing screenshots, without letting the agent issue arbitrary shell commands or silently fill missing parameters. It narrows natural-language intent to a single allowlisted command, such as node zuma.js ..., reducing unintended side effects.

How It Works

• Command mapping: after a trigger phrase, it only maps to predefined routes such as route-x-8, route-xiaohongshu-1, check-log, and take-screenshot.
• Explicit parameters: values like {username} and {interval} must be requested; obvious fixes may remove @, trim spaces, or extract minutes, while ranges and invalid characters block execution.
• Execution boundary: except for screenshots, it first runs node zuma.js setup; it executes one command at a time, disallows &&, pipes, and chaining, and returns raw errors without guessing repairs.
• Output contract: successful JSON is returned as-is, screenshots must include the image with an optional link, and failures use [ERROR]: ....

Limits

This skill fits fixed, template-driven operations, not open-ended system administration, arbitrary file operations, or autonomous troubleshooting. It depends on a local node environment and ZumaRobot.exe. If setup reports failure, execution stops and the original message is surfaced. Users should state clear intent, such as start versus stop or one-off versus scheduled, or conflict handling will interrupt the request.