More screenshots
Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+21 moreClaude CodeCursorCodex CLIMuseDots
Manus
Grok
Grok Bot+18 agentsNano Banana: Gemini Image Generation and Editing
Direct Gemini-class image models like an art director: six-slot prompts, wide-then-narrow iteration, reference chaining for consistent characters, and a scored QC rubric.
$5
Nano Banana: Gemini Image Generation and Editing
Example session with this skill installed
I need a consistent mascot for my coffee brand: a fox barista in three images — a clean logo-style pose, waving hello, and holding a takeaway cup — warm and friendly, for packaging and social posts.
- Read your context and instructions
- Compiled the nano banana
PROMPT RECORD, run by applying the skill:
Brief: mascot series of 3, packaging + social. Anchor first, series second.
Six slots, anchor prompt: Subject — "a fox barista standing upright, russet fur, cream apron over a green work shirt, friendly open expression, ears up." Medium — "flat vector-style illustration, clean shapes, minimal shading." Composition — "full body, centered, generous margin on all sides." Light — "even, bright, no dramatic shadows" (flat work doesn't need mood light; naming it stops the model adding some). Palette — "russet orange, cream, deep green, warm brown." Constraint — "plain cream background, no text, no props in paws."
Wide pass: 6 variants varying medium (flat vector vs. gouache vs. 3D render) and stance. Skeleton picked: the flat-vector front stance — it reads at thumbnail size, which is the real packaging test. Narrow pass: two single edits — apron pocket removed (it rendered as a smudge at small size), tail curled slightly for a friendlier silhouette.
Invariant locked verbatim for the series: "Same fox barista: russet fur, cream apron, green work shirt, ears up, flat vector illustration, russet/cream/deep-green palette, plain cream background, no text."
Series: waving pose (composition varied: three-quarter, one paw raised) and cup pose (paw holding a plain takeaway cup, added via the constraint slot as the single allowed prop). Trio check side by side: the cup image drifted the apron to beige — killed and regenerated with the palette slot restated. Rubric: identity 2, medium 2, palette 2 after the fix, variety 2, cleanliness 2 = 10/10. Three images, one fox.
Connects securely to your tools. The creator never sees your data.
About this skill
Image models fill every slot you leave blank with their own defaults, and defaults are where generic output comes from. This skill treats prompting as art direction. Every prompt is architected in six slots — subject, medium, composition, light, palette, constraint — and every generation is logged in a prompt record so a winning image can be reproduced or fixed later. Iteration runs wide-then-narrow: a broad pass that varies medium and composition to find a strong skeleton, then small named edits, one at a time, with a stop rule and a restart-or-repair decision point so you never pay for a fifth fix on a broken base. For series work, a locked anchor reference plus a verbatim invariant block in every prompt holds characters steady, checked with a generate-three-and-kill-the-drift rule and a five-dimension readiness rubric. Editing gets its own protocol (anchor clause first, one change per edit, mask for surgical fixes, always branch from the best copy), and a diagnosis table maps each failure symptom to the one slot that caused it.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Video Skills.
- It directs the model; it doesn't add model capabilities. Fine lettering in images stays unreliable — the skill routes real text to a design tool instead of promising clean type.
- Very long series (10+ images) still drift; the protocol manages it with re-anchoring on the original reference every 4–5 images, but per-image QC remains manual.
- Deliverables are generated raster images. If you need editable vector source files or print production art, that remains an illustrator's job.
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 buying a skill.
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 18 days ago
Permissions
No special permissions declared or detected