git:20260218.d9d797d to git:20260219.310d5fa

---

name: infographic-v2

description: Generate professional infographics using Nano Banana MCP (Gemini AI image generation). Follows a guided flow - analyze content, suggest visualizable concepts, propose visualization approaches, then generate on-brand images. USE THIS SKILL WHEN user says "create infographic v2", "make a visual v2", "infographic-v2". Modular architecture with lazy-loaded reference files for each phase.

---

# Infographic Generator

Create professional infographics in the BenAI brand style for LinkedIn and newsletters. Uses Gemini AI via the Nano Banana MCP server for image generation.

**Connectors:** Nano Banana MCP (Gemini AI image generation)

**Core philosophy:** The quality of the output is determined by the quality of the thinking BEFORE any prompt is written. Spend 70% of the effort understanding what to visualize and 30% on how to visualize it.

---

## When This Skill Loads

**DO NOT load any reference files yet.** Files load lazily, one phase at a time.

1. Execute Phase 0 (silent config check)

2. If API key is NOT configured: skip to Phase 5 (API Key Setup). Do NOT ask about content.

3. If API key IS configured: begin Phase 1.

---

# UNDERSTAND

## Phase 0: Silent Config Check

Run silently. Do NOT ask questions yet.

### 0.1: Ensure Directory Structure

```bash

mkdir -p .infographic/images

mkdir -p .infographic/prompts

```

### 0.2: Check for API Key

```bash

if [ -n "$GEMINI_API_KEY" ]; then

echo "API key found in environment"

elif [ -f .env ] && grep -q "^GEMINI_API_KEY=.\+" .env; then

echo "API key found in .env file"

else

echo "API key not configured"

fi

```

**Decision:**

- Found (env var OR .env): proceed silently to Phase 1

- NOT found: jump IMMEDIATELY to Phase 5 (API Key Setup)

### 0.3: Check for Brand Config

```bash

if [ -f .infographic/brand.md ]; then

cat .infographic/brand.md

fi

```

If brand config exists, acknowledge briefly: "I found your saved settings. Ready to create your infographic!"

---

## Phase 1: Content Intake

**No reference files loaded.** Just collect the raw material.

---

```

>>> READ references/content-analysis.md NOW

```

Apply the framework from `content-analysis.md`. Extract meaning at all 4 layers:

- Layer 1 (Narrative): What happened?

- Layer 2 (Themes): What is it about?

- Layer 3 (Claims): What does it SAY is true about the world?

- Layer 4 (Information): What concrete data, comparisons, or frameworks exist?

Present the analysis explicitly labeled by layer with honest strength ratings (Weak / Moderate / Strong).

```

```

---

# PLAN

```

options:

```

```

```

---

# EXECUTE

```

```

1. **Text inventory:** List every piece of text that must appear in the image.

Will use TEXT LINE N: format with CRITICAL exact spelling instruction.

2. **Element control:** Should icons appear? If NO, will use the nuclear NO formula

(listing every synonym: icons, illustrations, pictograms, emojis, visual symbols).

If YES, which icon library reference? (Phosphor, Feather, etc.)

3. **Style anchor:** What single brand/product reference captures the target aesthetic?

(e.g., "Stripe.com", "Linear.app", "Notion.so")

4. **Layout structure:** How many sections? Will use === SECTION === markers

with [LARGE GAP] between them for clear visual separation.

5. **Aspect ratio:** Confirm ratio appears in BOTH the prompt text AND the

generate_image parameter.

Use the correct architecture from `prompt-engineering.md`:

**For Editorial Illustration:** 9-part structure

Style anchor, emotional register, the scene, spatial composition, texture/materiality, color palette, text elements, what not to do, technical specs.

**For Information Graphic:** 8-part structure

Format declaration, information hierarchy, data visualization, exact text, brand system (from `brand-guidelines.md`), density control, scanability, technical specs.

For the selected platform, use the matching aspect ratio:

| Platform | aspectRatio param |

|----------|------------------|

| LinkedIn | "4:5" |

| Instagram (square) | "1:1" |

| Twitter/Presentation | "16:9" |

```

Call tool: generate_image

```

```bash

mkdir -p .infographic/images

cp "./generated_imgs/[returned-filename]" ".infographic/images/[topic-slug]-v1.png"

```

**IMMEDIATELY display to user using Read tool:**

```

Read file: .infographic/images/[topic-slug]-v1.png

```

The user MUST see the image to give feedback.

---

```

options:

```

**First edit:**

```

Call tool: edit_image

Parameters: {

"imagePath": ".infographic/images/[topic]-v1.png",

"prompt": "Edit this infographic:\n1. [Change 1]\n2. [Change 2]\nKeep everything else exactly the same."

}

```

**Subsequent edits:**

```

Call tool: continue_editing

Parameters: { "prompt": "[further changes]" }

```

After each edit:

1. Copy from `./generated_imgs/[filename]` to `.infographic/images/[topic]-v[N].png`

2. Display using Read tool

---

```

```

---

## Progressive Updates

When a user provides feedback that constitutes a general preference (not a one-time edit), the skill should recognize it as a potential rule.

Use `AskUserQuestion`:

```

question: "Should I remember this as a permanent rule for future infographics?"

header: "New Rule"

options:

- label: "Yes, always"

description: "Add this to my permanent rules"

- label: "Just this time"

description: "Apply it now but don't save"

```

If confirmed: append a new numbered rule to the Rules section below.

---

## Reference File Load Map

| Phase | File Loaded | Purpose |

|-------|------------|---------|

| 0 | None | Just checking config |

| 1 | None | Just collecting content |

| 2 | `content-analysis.md` | 4-layer framework for analysis |

---

## File Naming

| Stage | Pattern | Example |

|-------|---------|---------|

| First version | `[topic]-v1.png` | `chosen-pain-v1.png` |

| After edits | `[topic]-v2.png` | `chosen-pain-v2.png` |

| Final | `[topic]-final.png` | `chosen-pain-final.png` |

| Series | `[topic]-01-v1.png` | `tips-01-v1.png` |

| Prompts | `[topic]-prompt.md` | `.infographic/prompts/tips-prompt.md` |

---

## Directory Structure

```

.infographic/

brand.md # Saved brand config (optional, for overrides)

images/ # All generated infographics

topic-v1.png

topic-v2.png

topic-final.png

prompts/ # Saved prompts (for no-key fallback)

topic-prompt.md

```

---

## Rules

1. Never use em dashes (the long dash character)

2. Always give 5+ suggestions at every decision point (never one-off outputs)

3. Visualize claims (Layer 3) or information (Layer 4), NEVER narrative or themes

5. Edit don't regenerate (unless the concept itself is broken)

6. Use style anchors, not abstract descriptors ("Moebius meets Adult Swim" beats "clean modern illustration")

7. NEVER use black backgrounds

8. Max 4-5 distinct elements per infographic, one key message rule

9. Always include footer: "Ben Van Sprundel | Founder @ BenAI"

10. Instruct clearly in the process when to use each reference file (lazy loading)