wuyoscar/GPT-Image2-Skill / skills/get-prompt-from-image
get-prompt-from-image
VerifiedAnalyze user-provided reference images and reverse-engineer high-fidelity AI image-generation prompts. Use when the user asks to recreate, imitate, reverse-engineer, or extract prompts from photographs, illustrations, 3D renders, products, characters, landscapes, typography, logos, posters, or other visual references. Do not use for requests that only require OCR or an ordinary image description.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 7.3 KB |
| agents/openai.yaml | config | 279 B |
| references/analysis-framework.md | markdown | 6.1 KB |
| references/category-guides.md | markdown | 7.7 KB |
| references/illustration-style.md | markdown | 5.5 KB |
references/category-guides.md
View on GitHub# Subject and Medium-Specific Analysis
Select the relevant sections according to the image’s actual type. Images with multiple subjects may combine multiple sections, but must not force in absent features.
## General Image
Cover the subject content, artistic or medium style, color scheme, lighting, composition logic, material qualities, camera language, spatial layers, mood, image-quality requirements, and detail treatment. The Chinese and English positive Prompts must have the same meaning; do not add new elements in the English version.
## People and Portrait Photography
Analyze clearly visible age cues and gender presentation, facial contour, features, hairstyle, makeup, expression, gaze, head angle, body orientation, posture, shoulder and neck position, hand and leg placement, body weight, the body part closest to the camera, perspective emphasis, clothing, accessories, skin texture, and the relationship between the person and the environment. When identity, age, or ethnicity cannot be judged reliably, describe only visible appearance and do not infer identity.
## Photography
Analyze natural, hard, soft, back, or studio light; depth of field and focus; wide-angle, natural-perspective, or telephoto-compression effects; cool/warm, vintage, commercial clarity, or film character; clarity, grain, noise, sharpness, and motion blur; and documentary, studio, casual snapshot, fashion portrait, or cinematic-still photography. Write specific camera parameters only with sufficient evidence; do not invent focal length, aperture, shutter speed, ISO, camera model, or photographer name.
## Product and Commercial Photography
Analyze product category, shape, color, orientation, share of frame, packaging structure, edges, buttons, ports, label position, surface material, specular highlights, roughness, transparency, shadow placement, props such as droplets/powder/liquid/steam/ice/petals, background arrangement, and which visible structures and materials communicate the product’s selling points. Do not generate clear brand logos, garbled packaging text, or nonexistent functional structures.
## Typography, Logos, and Text Design
Analyze modern, vintage, cyber, handwritten, minimal, or decorative style; letterform width, slant, roundness, geometry, and proportions; stroke weight, serif or sans-serif structure, bevels, cuts, connections, broken strokes, and negative space; letter spacing, line spacing, hierarchy, and layout logic; outlines, glow, gradients, embossing, shadows, and 3D effects; metal, matte, glass, plastic, neon, or paper materials; and the relationship among palette, composition, lighting, and background.
When a logo or text is a supporting element, translate it into an abstract symbol, simplified pattern, or blurred mark. Write specific text in the Prompt only when the user explicitly requests it and provides the content.
## Natural Landscapes and City Scenes
Analyze location and environment type, season, weather, dawn/day/dusk/blue hour/night, foreground/middle ground/background and horizon, sky/clouds/water/mist/atmospheric perspective, vegetation/mountains/buildings/streets/lights/reflections, street perspective lines and building geometry, light direction, warm/cool changes, depth of field, perspective, detail texture, and overall mood. Do not invent a specific location, city, or landmark from a vague visual effect.
## Architecture and Interior
Analyze space type and function, perspective direction, vanishing points, main lines, foreground/middle ground/background, architecture/furniture/wall/floor/decorative materials, natural/window/strip/ambient light sources, spatial scale, order and mood, whether the space should be tidy/empty/free of clutter, and the connection between interior and exterior or the view outside the windows.
## Illustration
When the image’s primary medium is hand-drawn, digital-painted, flat-shaded, painterly, watercolor, pencil, printmaking, vector, anime-style, children’s-book, or another illustration, continue reading and applying [illustration-style.md](illustration-style.md). This section separates subject content from style expression and prioritizes observable techniques such as line, shape, value, color, and texture. Do not use it to rewrite analysis rules for photography, 3D renders, product images, or other non-illustration types.
## 3D Rendering
Analyze realistic, semi-realistic, 3D cartoon, clay, toy, or minimalist style; modeling proportions, soft edges, and detail precision; matte, metal, glass, acrylic, resin, PVC, or ceramic materials; reflection, refraction, roughness, subsurface scattering, and PBR qualities; three-point, soft, rim, volumetric, or ambient light; shadow softness, contact shadows, antialiasing, and post-processing glow. Write C4D, Blender, Octane, or other software or renderer names only with sufficient evidence or when specified by the user; otherwise use general 3D-rendering language.
## IP Characters, Chibi, Collectible Toys, and Blind Boxes
Analyze chibi, collectible-toy, comforting, traditional-inspired, clay, or cartoon style; head-to-body ratio, body shape, silhouette, facial features, expression, hairstyle, demeanor, clothing, decorations, action, matte/resin/PVC/ceramic/clay qualities, soft light, solid-color background, display stand, product-like presentation, full-body design, and detail density. The final Prompt should describe the character’s visual features without depending on a specific IP or character name.
## Animals
Analyze animal type, body shape, fur/feather/scale color, posture, action, gaze, fur/feather/scale/skin/claw texture, relationship with the environment or other subjects, rim light, and overall mood. When the species cannot be identified reliably, describe visible posture and appearance without inventing a specific breed.
## Food
Analyze food type, main ingredients, shape, color, plating, share of frame, oily sheen, sauce, steam, crispness, moisture, cream, powder, frosting, stretch, drips, tableware, tabletop material, lighting, shadows, depth of field, shooting angle, and which colors, gloss, textures, and sense of temperature create appetite appeal.
## Vehicles
Analyze vehicle type, viewing angle, body or fuselage material, reflections, lights, highlight lines, stationary or moving state, road/sky/city/natural environment, and actually present effects such as speed lines, dust, splashes, or motion blur, as well as perspective and subject scale. By default exclude clear brand logos, license plates, numbers, and readable text.
## Plants and Flowers
Analyze plant type or visible form, color, quantity, growth state, petal/leaf/pollen/dew/texture, whether light passes through petals or leaves, foreground and background occlusion, blur, and the relationship with the background, as well as the concrete visual sources of softness, freshness, romance, or quietness.
## Graphic Posters and Visual Design
Analyze layout structure, main-visual position, color blocks, graphic elements, background relationships, negative space, grid, alignment, visual flow, foreground/background hierarchy, occlusion, proportions, text-area position and hierarchy, main/supporting/accent colors, grain, gradients, outlines, shadows, collage, or glitch effects. If the goal is to reproduce the layout rather than the text content, write “reserved text area”; if text is only a visual element, exclude clear readable text.
## Abstract Graphics and Mixed Subjects
Analyze basic geometry, curves, color blocks, textures, repetition, primary/secondary relationships, rhythm, balance, directionality, transparent overlays, glow, gradients, materials, and the distance, occlusion, proportions, and narrative relationships among multiple subjects. Combine only the specialized rules that are actually relevant; avoid writing every category’s features into the Prompt at once.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Unverified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |