backend-login
Skill1 starsChanged 23 days agoMITView on GitHub
Playbook for sign-in and the backend: Google login or email login, Supabase databases and permissions, payments (Stripe, PayPal), sending email, connecting a custom domain, DNS, deploying to Vercel, Cloudflare or Hostinger, and Google Apps Script. Includes step-by-step console guides with exact clicks, the pop-ups to expect, and checks after each step. Use it whenever the user needs people to log in, data saved, payments taken, emails sent, a site put live on their domain, or says "set up the backend", "add Google login", "connect my domain", "why can't I log in", "deploy this".
- Reads credentials
In this repository7 other files it gives its agents.
- appSkill
- builder-setupSkill
- buildSkill
- project-memorySkill
- researchSkill
- voice-agentSkill
- websiteSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 36
Read back after each step: ask "what do you see now?", or verify it yourself through the API or CLI.
- Very specific: line 37
If a step is something you can do through a CLI or API with their logged-in tools (Vercel CLI, Supabase CLI, gh), do it yourself and tell …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Login, backend and going live
These steps cost the most time in the source data, because most of them happen in websites only the user can click:
- a CRM admin sign-up took 9 hours, because the user was pointed at a screen he couldn't see
- a Google sign-in quoted as "about 5 minutes" took 1 hour 50, because the guide skipped what Google requires
- a database column sat un-added for 3 days, buried inside reports
The fix is planning the user's steps up front and walking them through each one with exact clicks.
<kit> = the Builder Kit folder named in the session-start notice (usually ~/.claude/skills/builder-kit). Console guides live in consoles/.
Before any setup
- Final domain first. Login, payments and email all tie to a domain. Ask: "Will this live on your own domain? Is it connected yet?" If not, build everything else and batch the console steps for when the domain is ready. Setting up Google sign-in on a temporary address caused 2 of 4 corrections in one build, and the brand check couldn't pass until the move.
- Who owns the accounts? The user's Google Cloud, Supabase and domain. For client work, the client's, with the user invited. Collect this on day one: one site assumed the client had hosting, and the account had zero hosting plans.
- Run the backend preflight (
app/references/backend-preflight.md) if a database is involved. - Write the YOUR HANDS list on the Task Card: every console step with its guide, its URL, roughly how long it takes, and which stage it blocks.
Walking the user through a console
- Re-check the console against its live docs before giving steps (
research.mjs get <docs-url>, or web search for the current layout). Screens change, and guides written from memory were where time got lost. - Give one screen at a time: the page URL to open in their own browser, what to click, what to type (as a copy-paste block), the pop-ups to expect and what to do with them, and what they should see when it worked.
- Never point them at a hidden or embedded browser they can't see. Give a normal link.
- Secrets never go in chat. For a client secret or API key, create a local file (for example
.env.local, already in.gitignore), open it for them in Notepad or TextEdit, and let them paste and save. Check it worked without printing the value. - Read back after each step: ask "what do you see now?", or verify it yourself through the API or CLI. Then move on.
- If a step is something you can do through a CLI or API with their logged-in tools (Vercel CLI, Supabase CLI,
gh), do it yourself and tell them what you did.
Guides (in consoles/)
| Task | Guide |
| Google sign-in (OAuth client, consent screen, branding, publishing) | google-oauth.md |
| Database, sign-in providers, permissions | supabase.md |
| Deploy and connect a domain on Vercel | vercel-domain.md |
| DNS at Cloudflare (records, proxy, email records) | cloudflare-dns.md |
| Hostinger (domain, DNS, hosting plan check) | hostinger.md |
| Google Sheets as a lead inbox via Apps Script | apps-script.md |
| Payments (Stripe or PayPal: test mode first, then live) | payments.md |
| Sending email (Resend, SMTP, domain records) | email-sending.md |
| AI phone agent: ElevenLabs agents + a Telnyx or Twilio number | elevenlabs-telnyx.md |
| WhatsApp Business for an agent (WhatsApp Cloud API) | meta-whatsapp.md |
Prove it
For every kind of user (new, returning, already signed in as owner, wrong account), sign in on the live domain. Check that the data saves and that permissions hold (a stranger sees nothing). Test payments in test mode end to end before going live. If Claude can't do a real sign-in itself, give the user a three-line test script and ask for the result before saying done.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

