- Home
- Skills
- Image & Video
- Imagen: Gemini Image Generation Workflows
More screenshots
Works with the AI Agents you already use
Imagen: Gemini Image Generation Workflows
Direct Imagen on Gemini like an illustrator, not a slot machine: four-part prompts, three judged rounds, and a consistency system for image sets that match.
$5
Imagen: Gemini Image Generation Workflows
Example session with this skill installed
Our SaaS landing page needs a hero image: something that says "calm control over a messy workflow," wide format, matching our indigo-and-bone brand. We keep getting generic glowing-dashboard art. Give us the prompt and the process.
- Read your context and instructions
- Compiled the imagen
Brief first: the image sits above the fold and must make a visitor feel "my chaos, handled" in two seconds — so the subject is not a dashboard, it is a scene. Prompt, assembled in the four parts: "A tidy harbor office desk seen from above at dawn, scattered paper charts being aligned into one neat stack by a pair of hands, a brass compass and a mug at the edges (subject). Overhead flat-lay, subject centered slightly low, generous empty space in the upper third for headline text (composition). Editorial photography, muted indigo and bone palette, soft window light, shallow depth of field at the edges (style). 16:9, no screens, no text, no logos, no neon (constraints)."
Round 1 judges the idea only: desk, alignment, calm — right family, ship it forward. Round 2 locks the look: if Imagen drifts toward a stock-photo office, the fix is vocabulary, not adjectives — "harbor office desk, paper charts, brass instruments" replaces generic desk language, and the palette is named explicitly so the indigo survives. Round 3 makes it usable: the empty upper third is verified for the headline, edges checked for stray objects the crop would catch, and exclusions re-checked (one generated variant hid a glowing monitor in the corner — rejected on constraint compliance, not taste).
The deliverable is the prompt plus the accepted image and its log entry, so the blog header next month can reuse the same style block verbatim and stay in the same visual family.
Connects securely to your tools. The creator never sees your data.
About this skill
Most generated images disappoint for a boring reason: the prompt was a wish, not a direction. This skill treats Imagen the way an art director treats an illustrator who cannot see your mind. It opens with a one-sentence brief that fixes the image's job before any prompt is written, then assembles every prompt in four parts — precise subject, explicit composition, style named as a medium or movement with the palette stated, and constraints covering aspect ratio and exclusions. Generation runs in three rounds with separate judgments: round 1 judges only the idea, round 2 locks the look, round 3 makes it usable — and a stop rule sends you back to the concept instead of into a fourth polishing round when the idea itself is wrong. For sets of images, the skill builds a consistency system: a style block reused verbatim, an explicitly fixed palette, reference-first discipline for recurring characters, and a lineup test that makes drift visible, backed by a drift-diagnosis table. It ships with a fill-in prompt template, a pre-generation checklist, a five-dimension acceptance rubric (ship at 8+/10, no zero on brief fit), a complete worked example producing a six-illustration documentation set in one afternoon, and a failure-mode table covering gibberish text, centered-blob layouts, uncanny people, and brand mismatch — each with the actual fix. Practical plumbing is included: descriptive filenames, largest-size generation, prompt logging so every image stays reproducible, and API-key hygiene for Gemini access.
- It directs generation; it does not edit photographs or guarantee a specific output from any single prompt.
- Rendered text, logos, and real people or places are out of scope by design — the skill routes those to post-production or photography.
- Cross-image consistency is greatly improved by the style-block system, not made perfect; character-critical work still needs human review.
- Palette and style vocabulary transfer across generators, but round counts and drift behavior are tuned to Imagen on Gemini.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 24 days ago
Permissions
No special permissions declared or detected