content-images

name	content-images
description	Generate brand-consistent images for blog covers, Instagram posts, and carousels via AI (Gemini Imagen, Pollinations.ai fallback). Reads brand context from .octopus/content-images.json and GEMINI_API_KEY from .env.octopus. Caches by output path; --force overrides.
triggers	{"paths":[],"keywords":["og image","blog cover","instagram image","carousel","content image","imagem","capa","carrossel"],"tools":[]}

Content Images Protocol

Overview

This skill generates brand-consistent images for content campaigns. It resolves the brand preset, infers the output format from the briefing, checks the file cache, and generates via Gemini Imagen (or Pollinations.ai when no key is available). All content and prompts are dynamic per interaction — there are no hardcoded image sets.

The skill composes with the social-media agent: after the social-media agent finishes Phase 4 (Approval Gate), it asks whether to generate the visual assets and then dispatches to this skill with the campaign briefing.

Invocation

/octopus:content-images <briefing> [--force]

Arguments:

• <briefing> — conversational description of the image needed: topic, channel, intended filename or slug (e.g. "create an OG image for the article about student dropout, slug og-evasao")
• --force — regenerate even if the output file already exists

Execution Flow

Execute in order for each image in the briefing:

Step 1 — Load brand preset

cat .octopus/content-images.json 2>/dev/null

If the file exists, extract:

• brand.name — project name
• brand.visual_style — aesthetic descriptors
• brand.subjects — main visual subjects
• brand.negative — negative constraints (default: "no text, no logos, no watermarks")
• outputs.og — output dir for OG images (default: ./public/blog/)
• outputs.instagram — output dir for Instagram images (default: ./public/social/)

If the file does not exist, ask the user for brand context interactively before proceeding.

Step 2 — Parse briefing

From the briefing, determine:

• topic — what the image should depict
• format — og, instagram, or carousel (see Format Inference below)
• slug — filename without extension (e.g. og-evasao-alunos)
• slide count — for carousel only (infer from briefing; default: 5)

Step 3 — Resolve output path

og         →  <outputs.og>/<slug>.jpg
instagram  →  <outputs.instagram>/<slug>.jpg
carousel   →  <outputs.instagram>/<slug>/slide-NN.jpg  (for each slide)

Step 4 — Cache check

test -f <output-path>

• If file exists and --force is not set: print ✓ already exists: <path> and skip.
• If file does not exist or --force is set: proceed to generation.

Step 5 — Resolve API key

# Try environment first, then .env.octopus
echo "${GEMINI_API_KEY:-$(grep '^GEMINI_API_KEY=' .env.octopus 2>/dev/null | cut -d= -f2-)}"

• If a non-empty key is found → use Gemini Imagen
• If empty → fall back to Pollinations.ai; inform the user once

Step 6 — Build prompt

Professional <format-description> for <brand.name>.
Topic: <topic>.
Style: <brand.visual_style>.
Subjects: <brand.subjects>.
<brand.negative>.
<format-constraints>

Format-specific constraints appended:

• og → "16:9 widescreen aspect ratio, 1200×630 pixels, blog cover composition."
• instagram → "Square 1:1 aspect ratio, 1080×1080 pixels, Instagram feed composition."
• carousel → "Square 1:1 aspect ratio, 1080×1080 pixels, slide <N> of <total>: <slide-topic>."

Step 7 — Generate image

With Gemini key:

// node -e inline script
const { GoogleGenAI } = require('@google/genai');
const fs = require('fs');

const ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });
const response = await ai.models.generateImages({
  model: 'imagen-3.0-generate-002',
  prompt: '<prompt>',
  config: { numberOfImages: 1, aspectRatio: '<ratio>' },
});
const base64 = response.generatedImages[0].image.imageBytes;
fs.writeFileSync('<tmp-path>.png', Buffer.from(base64, 'base64'));

Aspect ratio values: 16:9 for OG, 1:1 for Instagram/carousel.

Without Gemini key (Pollinations.ai fallback):

curl -sL "https://image.pollinations.ai/prompt/<url-encoded-prompt>?width=<w>&height=<h>&nologo=true" \
  -o <tmp-path>.png

Width/height by format: 1200×630 (OG), 1080×1080 (Instagram/carousel).

Step 8 — Process and save

// node -e inline script using sharp
const sharp = require('sharp');
await sharp('<tmp-path>.png')
  .resize(<width>, <height>)
  .jpeg({ quality: 85, progressive: true, mozjpeg: true })
  .toFile('<output-path>');
fs.unlinkSync('<tmp-path>.png');

If sharp is not installed, save the PNG directly to the output path and inform the user that JPEG conversion requires npm install sharp.

Step 9 — Report

✓ generated: <output-path>  (via gemini | pollinations)

For carousels, report each slide on its own line.

Format Inference

Infer the output format from these signals (first match wins):

Signal in briefing	Format
"blog", "artigo", "post", "og", "open graph", "capa do post"	og
"carrossel", "carousel", "slides", "série de"	carousel
"instagram", "feed", "square", "quadrado", "post ig"	instagram
none of the above	ask the user to clarify

Supported Formats

Format	Dimensions	Use
og	1200 × 630 px	Blog covers, Open Graph meta tags
instagram	1080 × 1080 px	Instagram feed posts
carousel	1080 × 1080 px × N	Instagram carousels (N slides)

API Key Setup

GEMINI_API_KEY is read using this priority order:

1. $GEMINI_API_KEY environment variable (already exported in shell)
2. .env.octopus file at the project root

Required .env.octopus entry:

GEMINI_API_KEY=your-google-ai-studio-key

Never print or log the key value. If both sources are empty, fall back to Pollinations.ai and report: ⚠ GEMINI_API_KEY not found — using Pollinations.ai (lower quality).

Brand Preset Setup

Create .octopus/content-images.json once per project and commit it:

{
  "brand": {
    "name": "Your Project Name",
    "visual_style": "dark, moody, editorial, cinematic",
    "subjects": "describe the main visual subjects of your brand",
    "negative": "no text, no logos, no watermarks"
  },
  "outputs": {
    "og": "./public/blog/",
    "instagram": "./public/social/"
  }
}

Dependencies

Ensure these npm packages are installed in the project:

npm install @google/genai sharp

@google/genai is required for Gemini. sharp is required for JPEG conversion and resizing. Both are optional if using Pollinations.ai and PNG output, but strongly recommended.

Examples

/octopus:content-images create an OG image for the article about reducing student dropout, slug og-evasao-alunos

/octopus:content-images generate 5 Instagram carousel slides about the benefits of automated billing for martial arts academies

/octopus:content-images regenerate the feed post for the summer campaign --force