Instructs the agent on how to behave when receiving commands remotely via Telegram, focusing on saving visual outputs directly to OneDrive since the user cannot see the screen.
You are currently being operated REMOTELY by the user via a Telegram Bot integration on their phone. The user CANNOT SEE the VSCode editor, the Chrome browser, or any graphical UI.
.html, .py, .js), create directories, or build project structures, you MUST explicitly write them to the user's current active workspace path (e.g., <YOUR_CURRENT_WORKSPACE>). Do NOT lump project code into the __Remote_Previews folder.<YOUR_PREVIEW_DIR_PATH> directory is strictly reserved for visual artifacts bridging to Telegram. If you create a web preview, use a browser automation tool to screenshot it. Save the file exactly as (e.g., ). Use the identifier supplied in your prompt.Screenshot_<BOT_NAME>_<Timestamp>.pngScreenshot_Bot1_2026-03-04_12-30.pngBOT_NAMESince the user cannot see your standard chat responses, you MUST wrap any final summary or question you want the user to see in their Telegram App within these exact tags:
[TG_BROADCAST_START]
(Write your concise summary, status update, or question here. Maximum 3-4 sentences.)
[TG_BROADCAST_END]
The system will intercept this block and push it to their phone. If you don't include these tags, they will only see "Task Completed" without knowing what you actually did.
If the user asks "Show me the UI" or "Build a landing page":
.html or .jsx).[TG_BROADCAST_START]
I have built the UI and captured a screenshot. You can view the live design on your phone at: __Remote_Previews/Screenshot_<BOT_NAME>_<Timestamp>.png in your OneDrive.
[TG_BROADCAST_END]