Engineering build desk
Code, review, and audit — you still merge.
6 bots · 30 min to stand up
Open blueprintAll use cases
Engineering
Tip only: webhook the login Grok Bot can finish
Coding agents stall when a human has to sign in.
What hurt
Henry’s published tip: Cursor pings a Grok Bot through a webhook with what the coding agent needs. The bot checks it is not dangerous, then signs up or signs in and sends the info back. That is the whole method. No shared prompt.
What to install or create
One login-helper bot. Routine trigger = webhook. Cursor POSTs the blocked step. Bot refuses anything dangerous, completes the browser login you already allowed, replies to the agent. You still own new accounts and passwords.
Bots
Login helper
Wakes on webhook. Checks the ask. Finishes a login you already allowed. Sends the result back.
Plugins
Browser
Required
Files
The cited tweet. Optional: a short allow-list of sites the helper may sign into.
Routines
Webhook trigger. Cursor POSTs the blocked sign-in. Helper replies. No schedule needed.
The prompt
This is a tip, not a pasteable system prompt. The author did not publish one. Follow the first-run steps.
First run
- 1Read the cited tweet. Tip only — no pasted prompt.
- 2Create a routine, add trigger, select webhook. Copy the URL/token into Cursor.
- 3Tell the bot: refuse dangerous asks; only finish logins on my allow-list; send the result back.
- 4When a coding agent stalls on sign-in, POST that step. Watch the first run.
- 5You still create accounts and type passwords. The bot does not spend or publish.
Approval fence
Do not create accounts, spend, or submit credentials unless you said yes. Refuse anything that looks dangerous.
What to tweak
Steal the split: agent codes, bot does the signed-in click, you still gate accounts.
Source
Henry @Henry_J_Brew · Original post
Not affiliated with xAI. We did not write this prompt. Third-party template terms still apply.