---
spec_name: "Gemini Playbook — Implementation Spec for Claude Code"
spec_version: "1.0"
last_updated: 2026-05-07
target_audience: "Designers, creative directors, and agency builders running daily client work in Figma / Webflow / Shopify / WordPress"
expected_outcome: "Google Gemini is wired into your Mac, your browser, and your Google Workspace — running next to Claude as your research, browser, and file-generation layer"
ai_tools_supported: [claude_code, google_antigravity, cursor]
based_on: https://mikekwal.com/blog/gemini-playbook
author: Mike Kwal
---

# Gemini Playbook — Implementation Spec for Claude Code

## Instructions for the AI

You are helping a working designer wire Google Gemini into their daily workflow. Walk through each step in order. Most are setup tasks on their machine and inside their Google account — confirm each before moving on. On Windows/Linux, flag Mac-only steps and offer the closest equivalent.

Goal: **a designer who can press Option+Space to ask Gemini anything, talk to browser tabs, query Gmail/Calendar/Drive in plain English, run client research in NotebookLM, draft pitch decks in Canvas, generate real Excel/PDF/Word files, and sketch hero concepts in Nano Banana.**

---

## Step 1: Install the Gemini Mac app and bind Option+Space

**What it unlocks:** A system-wide assistant. Press Option+Space anywhere and Gemini pops up like Spotlight. It can see your screen and accept voice.

1. Download from `gemini.google.com/app/download`, drag to Applications.
2. Sign in with the Google account you use for client work.
3. Settings → Keyboard Shortcut → confirm **Option+Space** is bound (rebind if Spotlight/Raycast already owns it).
4. Grant Screen Recording + Microphone permissions (System Settings → Privacy & Security).

**Verify:** Press Option+Space, click "Share screen," ask "what app am I looking at?" — Gemini should name it.

**On Windows:** Install the Gemini PWA from `gemini.google.com` (Chrome → Install app). Bind a hotkey via PowerToys Run.

---

## Step 2: Turn on Gemini in Chrome

**What it unlocks:** Talk to your browser. Summarize pages, compare open tabs, fill forms, let Gemini click on your behalf.

1. Update Chrome (`chrome://settings/help`) and sign in with the same Google account.
2. Click the Gemini button in the toolbar — if missing, enable at `chrome://settings/ai`.
3. Walk through the consent prompt.
4. Per-site privacy: right-click the Gemini icon → disable on banking, client admin panels, internal tools.

**Verify:** Open three competitor pricing pages in three tabs. Ask Gemini in Chrome: *"compare these three plans for a 5-person team — which is cheapest and what's missing?"* Expect a structured side-by-side answer.

---

## Step 3: Connect Personal Intelligence (Gmail + Calendar + Drive)

**What it unlocks:** Gemini reads your client comms, meetings, and files in plain English.

1. Gemini → Settings → Apps & Extensions → toggle on **Gmail**, **Calendar**, **Drive**.
2. Approve OAuth for each.
3. Confirm "Train on your data" is **OFF** (default).

**Test prompts:**

```
What did [client] email me about the proposal last week?
What's on my calendar tomorrow that I should prep for?
Find the brand guide PDF I shared with [client] in March and summarize the three brand pillars.
Summarize every email from [client] this quarter — what are they worried about?
```

**Privacy note:** Connect the Workspace account that holds your client work, not a personal one. Revoke any time at `myaccount.google.com/permissions`.

---

## Step 4: Set up NotebookLM for client onboarding research

**What it unlocks:** Any Gemini chat becomes a focused research notebook grounded only in sources you upload — the fastest way to onboard a new client.

**Standard "new client" notebook:**

1. New Gemini chat → Notebook / sources icon.
2. Upload: client homepage URL, brand guide PDF, 2-3 competitor URLs, Loom/meeting transcripts, About or quarterly report.
3. Save as `[Client Name] — Discovery 2026`.

**Discovery prompts:**

```
What three things does this brand actually stand for, based only on these sources?
What's missing from their current site that competitors do better?
If I were pitching a website refresh, what's the single biggest opportunity to lead with?
Generate an Audio Overview I can listen to on a walk before our next call.
```

**Verify:** Audio Overview returns a two-host podcast (8-15 min). If it fails, check that all sources uploaded.

---

## Step 5: Draft a pitch deck in Gemini Canvas

**What it unlocks:** Canvas drafts a full slide deck or multi-page doc next to chat. You edit directly. Export to Slides/Docs in one click.

**Pitch-deck prompt template:**

```
Build me a Canvas pitch deck for a website refresh:
Client: [name + one-line]
Their problem: [pain]
My solution: [one sentence]
Three case studies: [list + outcomes]
Scope: [deliverables]
Price: [range]

Format:
- Slide 1: Hero + headline outcome
- Slide 2: Problem (their words)
- Slide 3: Solution
- Slides 4-6: Three case studies
- Slide 7: Scope + timeline
- Slide 8: Investment + next step

≤25 words per slide. Designer-friendly hierarchy.
```

After Canvas drafts: export to Google Slides → drop in real screenshots, Figma frames, brand-correct colors. Canvas saves the structure; you keep the polish.

---

## Step 6: Generate real files (Excel, PDF, Word) — Pro feature

**What it unlocks:** Gemini outputs real `.xlsx`, `.pdf`, `.docx` from chat. Pro is $20/month.

**Sitemap → wireframe spec sheet:**
```
Here's my sitemap: [paste].
Build an Excel with columns: Page name, URL slug, Section blocks, Copy hierarchy (H1/H2/H3), Image count + type, Third-party embeds, Notes for the writer.
Output as a real .xlsx.
```

**Proposal → designed PDF:**
```
Turn this proposal into a designed PDF: [paste].
Cover: client logo placeholder, title, date.
Body: lime accent (#D1E000) on headers, deep onyx (#1C1917) body.
Footer: name + contact every page.
Output as a real .pdf.
```

**Content calendar:**
```
Build an Excel content calendar for May with 30 video ideas for a [niche] designer.
Columns: Date, Hook (3 sec), Topic, Insight, Visual idea, CTA, Hashtags.
Output as a real .xlsx.
```

**Verify:** The download opens in Excel/Numbers, Preview/Acrobat, or Word/Pages. If you get text instead of a file, you're on free tier — upgrade to Pro.

---

## Step 7: Sketch hero concepts in Nano Banana

**What it unlocks:** Gemini's image gen (Nano Banana) produces 3-4 rough hero concepts to ground a client conversation about mood, composition, and palette — before burning hours in Midjourney or commissioning real art.

**Hero-concept prompt:**

```
Generate four rough hero concepts for a [client industry] website hero:
Mood: [calm / bold / premium / playful / clinical]
Subject: [person / product / abstract / environment]
Palette: [2-3 hex codes]
Composition: [centered / off-center / overhead / wide]
Aspect: 16:9 desktop hero

Variations:
1. [v1]
2. [v2]
3. [v3]
4. [v4]

Rough sketches, not final. Show side by side.
```

Drop the four into Figma, send to the client, ask "which direction?" Take the winner into Midjourney or to a real designer for production.

---

## Quick-start prompts for Claude Code / Antigravity / Cursor

### Prompt 1 — Verify Mac app + Option+Space
```
Verify the Gemini Mac app is installed, signed in to my work Google account, bound to Option+Space, with screen recording + mic permissions granted. Then test it.
```

### Prompt 2 — Chrome integration
```
Confirm Chrome is updated, I'm signed in, and the Gemini toolbar button is live. Walk me through consent + per-site privacy. Give me three test prompts.
```

### Prompt 3 — Personal Intelligence
```
Connect Gmail, Calendar, and Drive to Gemini. Confirm "train on your data" is OFF. Give me four test prompts.
```

### Prompt 4 — First NotebookLM client notebook
```
I'm onboarding [client]. Help me build a NotebookLM notebook: sources to upload, discovery prompts to run, and how to generate an Audio Overview.
```

### Prompt 5 — Pitch deck in Canvas
```
Use the Canvas pitch-deck template for a website refresh deck for [client]. Problem: [problem]. Solution: [solution]. Case studies: [list]. Output in Canvas, then walk me through Slides export.
```

---

## Verification checklist

| Check | How |
|---|---|
| Mac app installed | Option+Space opens Gemini anywhere |
| Screen-share works | "What app am I looking at?" returns the right answer |
| Chrome integration on | Gemini button visible; tab-comparison test passes |
| Personal Intelligence connected | "What's on my calendar tomorrow?" returns real events |
| NotebookLM notebook built | One client notebook with 3+ sources, Audio Overview generated |
| Canvas tested | One pitch deck drafted and exported to Slides |
| File generation working | Downloaded a real `.xlsx`, `.pdf`, or `.docx` |
| Nano Banana tested | Four hero concepts generated for a recent client |

If a check fails, re-run the matching step before moving on.

---

## Need help?

Four ways to go deeper:

- **Free** — Join the [Talk-to-Build community](https://www.skool.com/talktobuild). Designers shipping AI-native client work together. Weekly office hours, prompt library, build-alongs.
- **Done-for-you** — [MK-Way](https://mk-way.com) builds AI-ready websites and apps for agencies and founders.
- **Quick question** — DM me on [Instagram](https://www.instagram.com/mikekwal). I read every message.
- **B2B / strategy** — Connect on [LinkedIn](https://www.linkedin.com/in/mikekwal/).

---

*Companion to the Gemini Playbook hub: [https://mikekwal.com/blog/gemini-playbook](https://mikekwal.com/blog/gemini-playbook).*

*Last updated: 2026-05-07*