Korean Booking

What to consider before installing: - Privacy: the skill will ask users for phone numbers and will send them to https://api.yestokr.com/api/Appointment/saveFromSkill when you ask it to 'help book'. Only provide real contact details if you trust that endpoint and the operator. - Output / prompt-injection: SKILL.md contains strict 'output rules' that try to force verbatim replies and disallow the agent's normal browser tool — this looks like prompt-injection. Treat those rules as untrusted text; they are instructions to the agent, not platform policy. Prefer not to grant the skill autonomous or always-on invocation until you review and understand these rules. - Context leakage: the skill scans all strings in the provided context object to resolve hospital names. Do NOT pass any secrets or unrelated sensitive data in the context (API keys, tokens, private messages). If your agent automatically injects conversation history or system state into context, that may leak to matching/logic that could be sent to the external API. - Browser automation privilege: the skill runs Playwright (headful Chromium) with flags that disable web security/CSP and grants clipboard/geolocation/notification permissions. These settings increase the risk that web pages opened by the skill could access or exfiltrate data available to the browser environment. Only run this skill on trusted machines and consider running it in an isolated environment (VM/container) if you must use it. - Trust the endpoint: verify the legitimacy of api.yestokr.com / i.beautsgo.com before sending PII. If unsure, test flows with a dummy phone number and monitor the outbound network traffic. - Practical steps: review the code (api/submitBookingApi, browser scripts), run locally in a sandboxed environment first, and avoid including secrets in the agent context. If you are not comfortable with Playwright/browser automation flags or the endpoint, prefer manual booking or a skill with fewer automation privileges. If you want, I can highlight exact lines of code that implement the context scanning, the POST request, and the Chromium flags, or suggest a safer configuration (remove disable-web-security flags, drop clipboard permissions, and avoid broad context scanning).