Copy as markdown[View .md](https://docs.funnelfizz.com/ai-agents/hand-it-to-your-agent "View the raw markdown for this page")[Open in Claude](https://claude.ai/new?q=Read%20https%3A%2F%2Fdocs.funnelfizz.com%2Fai-agents%2Fhand-it-to-your-agent.md%20and%20help%20me%20with%20this%20FunnelFizz%20topic%3A%20Hand%20it%20to%20your%20agent "Open this page in Claude with context")[Open in ChatGPT](https://chat.openai.com/?q=Read%20https%3A%2F%2Fdocs.funnelfizz.com%2Fai-agents%2Fhand-it-to-your-agent.md%20and%20help%20me%20with%20this%20FunnelFizz%20topic%3A%20Hand%20it%20to%20your%20agent "Open this page in ChatGPT with context")

# Hand it to your agent

The simplest possible flow: paste a prompt into your AI coding agent, and it handles signup, tracking, Stripe, providers, and your first automation.

## The prompt[​](#the-prompt "Direct link to The prompt")

Paste this verbatim into Claude Code, Cursor, Windsurf, ChatGPT, or any other AI coding agent:

```
Read https://docs.funnelfizz.com/skill.md and follow the skill exactly. Walk me through each

step. Stop and ask me before doing anything destructive or anything that requires my credentials.
```

That's it. The agent will:

1. Fetch [skill.md](https://docs.funnelfizz.com/skill) and parse it.
2. Ask you the preliminaries it needs to know (your domain, what framework you're on, whether you have Stripe set up).
3. Walk you through signup and onboarding.
4. Write the tracking snippet into your codebase correctly for your framework.
5. Walk you through creating a Stripe restricted key with the right seven scopes.
6. Help connect providers (X, YouTube, GSC, whatever you use).
7. Write `funnelfizz('identify', ...)` calls into your auth layer.
8. Instrument custom events in key places.
9. Build a starter trial welcome automation.

## What the agent sees[​](#what-the-agent-sees "Direct link to What the agent sees")

[skill.md](https://docs.funnelfizz.com/skill) is a \~2500-word self-contained runbook covering every step above. It's specifically designed for one-shot agent consumption — no cross-page dependencies, no navigation chrome. The agent reads it once and has everything it needs.

## What you still need to do[​](#what-you-still-need-to-do "Direct link to What you still need to do")

Some things your agent can't do for you:

* **Create the FunnelFizz account** — you have to sign up yourself at funnelfizz.com (agents generally don't handle email-verification flows well).
* **Create the Stripe restricted key** — you'll click through Stripe's UI. The agent will guide you but won't click buttons for you.
* **OAuth provider connects** — you'll sign into X / YouTube / LinkedIn / etc. yourself. The agent will kick off each one and wait.
* **Approve changes to your codebase** — the agent will propose code changes. You confirm.

Everything else — reading snippets, writing code, generating config, drafting email copy — the agent handles.

## Tips for best results[​](#tips-for-best-results "Direct link to Tips for best results")

### Give it your codebase context upfront[​](#give-it-your-codebase-context-upfront "Direct link to Give it your codebase context upfront")

```
This is a Next.js 14 App Router project with next-auth. I'm on the Hobby plan. I use Stripe

for billing. My domain is widgetspro.com, marketing is at the root, app is at

app.widgetspro.com.
```

Saves several rounds of back-and-forth.

### Tell it what "activation" means for your product[​](#tell-it-what-activation-means-for-your-product "Direct link to Tell it what \"activation\" means for your product")

The skill file mentions firing an `activation` custom event. Activation is product-specific. Tell the agent:

```
For my product, "activation" means the user created their first project and invited at least

one teammate. Please instrument those two events.
```

### Ask it to commit along the way[​](#ask-it-to-commit-along-the-way "Direct link to Ask it to commit along the way")

```
After each major change (tracking install, identify call, event instrumentation), make a commit

with a clear message. Don't bundle unrelated changes into one commit.
```

### Double-check the Stripe scopes[​](#double-check-the-stripe-scopes "Direct link to Double-check the Stripe scopes")

Stripe keys are the most common failure point. When the agent asks you to create the restricted key:

* Name it `FunnelFizz (read-only)`.

* Turn **Read** on for exactly these 7:

  <!-- -->

  1. Customers
  2. Charges and Refunds
  3. Events
  4. Products and Prices
  5. Invoices
  6. Subscriptions
  7. Balance

* Leave everything else **None**.

If FunnelFizz says "missing scopes," the validation tells you exactly which ones — go back to Stripe, edit the key, add them.

## Security[​](#security "Direct link to Security")

Your agent should NEVER:

* Commit your Stripe key to a repo.
* Send your key, OAuth tokens, or session cookies to any third-party service.
* Create Stripe accounts or sub-accounts on your behalf.

skill.md explicitly instructs agents on these boundaries. If you see your agent about to violate one of them, stop it and reset.

## If something goes wrong[​](#if-something-goes-wrong "Direct link to If something goes wrong")

Ask the agent:

```
That didn't work. Paste the error you got into the context and walk me through diagnosis.

Don't retry blindly — let's understand what's happening first.
```

The most common failures:

| Symptom                               | Likely cause                                                         |
| ------------------------------------- | -------------------------------------------------------------------- |
| Tracking snippet not verifying        | Meta tag outside `<head>`, subdomain mismatch, CSP blocking script   |
| Stripe key rejected                   | Missing one of 7 scopes — FunnelFizz tells you which                 |
| `identify` never fires                | Script not loaded on authenticated pages, or auth state check misses |
| Custom events not appearing in splits | Events fired before `identify`, so attributed to anonymous visitor   |

## For agents writing this kind of skill file themselves[​](#for-agents-writing-this-kind-of-skill-file-themselves "Direct link to For agents writing this kind of skill file themselves")

If you're an agent or developer wondering how to write a skill.md for your own product, the canonical template is ours: [https://docs.funnelfizz.com/skill.md](https://docs.funnelfizz.com/skill). Moltbook's version ([moltbook.com/skill.md](https://www.moltbook.com/skill.md)) is also excellent.

Key principles:

* **Self-contained.** No cross-page links the agent needs to fetch.
* **Ordered.** Steps must run in sequence; dependencies explicit.
* **Conservative.** Err on the side of asking the user before destructive actions.
* **Versioned.** Include a `Version:` line. Update when your product changes.

***

**Next:** [llms.txt convention](https://docs.funnelfizz.com/ai-agents/llms-txt.md) · [MCP & Context7](https://docs.funnelfizz.com/ai-agents/mcp-and-context7.md)