Update to Givon AI

{
  "description": "For a generation-ready image prompt: turn the visual goal and the chosen model's contract into a standalone prompt with subject, composition and light. Not needed for video, audio or answers without a prompt.",
  "included_files": [
    {
      "relative_path": "references/product-and-text.md",
      "size_in_bytes": 4029
    }
  ],
  "name": "image-prompt-craft",
  "skill_md_contents": "---\nname: image-prompt-craft\ndescription: \"For a generation-ready image prompt: turn the visual goal and the chosen model's contract into a standalone prompt with subject, composition and light. Not needed for video, audio or answers without a prompt.\"\nmetadata:\n  version: skill-render-ea6bd93365667f6d\n---\n\n# Professional image prompt preparation\n\nYour task is not to retell the request but to turn it into a controllable visual brief for the\nexact chosen model. The finished prompt has to make sense to the generator without the chat\nhistory and carry enough decisions to produce a coherent frame. Length by itself is neither a\nvirtue nor a flaw.\n\n## 1. First establish the actual result\n\nDetermine what exactly the user wants to get:\n\n- a new image from scratch;\n- a change to an attached image;\n- a new frame that keeps a character, product, logo or other object;\n- a transfer of only the style, composition, background or palette from a reference;\n- an advertising key visual, a product card, a poster, a portrait, an illustration, a UI or an\n  infographic;\n- a series of consistent images.\n\nThe user's explicit facts (subject, action, properties, relations, text, style, prohibitions)\nare fixed. Do not turn everything else into a questionnaire: choose consistent non-essential\ndetails yourself when they help the task. Ask a question only when, without the answer, the\nresult itself cannot be understood or a compatible input cannot be chosen.\n\n## 2. Read images as evidence, not as decoration\n\nGive each attached image one or more precise roles:\n\n- `source` — the image to be changed;\n- `identity / character` — the appearance and stable traits of a person or character;\n- `product / object` — the product's shape, color, material, packaging and markings;\n- `style` — artistic language, texture, palette, processing;\n- `composition` — placement of objects, shot size, perspective;\n- `background / location` — the surroundings;\n- `logo / text` — exact graphics or lettering.\n\nLook at the available pixels and use only the properties that belong to the declared role. A file\nname and an earlier prompt are auxiliary metadata, not a substitute for the visual content.\nIf the user sent a screenshot of an interface, separate the target object from the buttons, frames\nand interface text: the interface does not become part of the result unless the user asked for it.\n\nDo not mix roles. A camera-motion reference does not set a character's appearance; a style\nreference does not change the product; a source image does not mean that absolutely everything\nmust be kept.\n\n## 3. Follow the chosen model's contract\n\nUse only the confirmed inputs and capabilities of the chosen model. Do not guess them from the\nbrand, the name or a memory of the catalog. The chosen model's current prompting guidance owns model-specific syntax;\nits live contract owns available fields.\n\nFor product references, exact lettering and local edits, read `references/product-and-text.md`\nwhen needed. It holds general craft, checked against the official GPT Image 2 and Seedream 5 Pro\nguides. Do not load it for a simple task where the current context is already enough.\n\nPass the aspect ratio, size, quality mode, number of results and other execution controls in the\ntyped settings. Repeat them in the prompt only when the value is itself part of the composition,\nfor example \"a vertical poster with room for a headline at the top\".\n\n## 4. Build an internal visual brief\n\nBefore writing the prompt, sort the decisions into four groups in your head:\n\n1. `keep` — all of the user's explicit facts and the invariants of the references;\n2. `change / create` — the observable result;\n3. `professional choices` — the angle, composition, setting, light, palette, material, texture\n   and depth of field that you choose;\n4. `do not add` — new characters, products, story facts, advertising promises, logos, lettering\n   or decor that change the meaning of the task.\n\nProfessional choices are needed when they make the result specific. They must not contradict\nexplicit values. Do not add empty boosters such as `masterpiece`, `best quality`, `8K`, `HDR` or\na random lens unless they express the required look.\n\n## 5. Architecture of a finished prompt\n\nDo not copy this list mechanically and do not output empty sections. For an ordinary generation,\nassemble a natural, coherent brief in this order:\n\n1. **Purpose and main frame** — what is being created and for what, if that affects the\n   composition.\n2. **Subject** — who or what is in the frame, and its important visible traits.\n3. **Action / pose / relations** — what the subject does and how the objects are placed relative\n   to each other.\n4. **Composition** — shot size, angle, perspective, focal point, negative space.\n5. **Scene** — the environment, the background and only the details that carry meaning.\n6. **Light and color** — source, direction, hardness, temperature, contrast, palette.\n7. **Material and texture** — how the surfaces look and how light behaves on them.\n8. **Visual language** — photography, CGI, anime, graphics, collage, and its specific traits.\n9. **Text and invariants** — verbatim lettering, logo, geometry or identity, when needed.\n\n### Creating a new image\n\nDescribe the final frame, not the agent's working process. Do not write \"generate\", \"use the\nmodel\" or \"choose the settings\" inside the prompt. One paragraph is fine if it stays readable; a\ncomplex commercial frame can be split into short blocks of meaning.\n\n### Editing a source image\n\nUse a clear logic:\n\n```text\nSOURCE: [which asset is being edited]\nCHANGE: [one or more specific changes]\nKEEP: [only the properties that truly must not change]\nLOOK: [what the final visual look should be]\nCONSTRAINTS: [side changes that are not allowed]\n```\n\nDo not list the whole image in `KEEP`. Pin down what the model can actually break by accident:\nidentity, geometry, angle, product color, logo, background or lighting.\n\n### Person, avatar or character\n\nIf a new look is needed without an identity reference, describe the age group, visible features,\nhair, clothing, expression, pose, gaze and light only as far as the task needs. If there is an\nidentity reference, do not replace it with a long invented description: state the reference's\nrole and pin down only the critical traits and the allowed change.\n\n### Product and advertising\n\nThe product stays the visual anchor. Fix its shape, color, material, packaging and logo from the\nreference, then choose a commercially meaningful composition, setting, light, contact with the\nsurface, shadows and room for copy. Do not invent product properties or marketing claims.\n\n### Lettering and infographics\n\nPass the user's text verbatim in quotes. State the hierarchy, placement, typeface character,\ncolor, contrast and free area. Do not correct the brand, spelling or letter case unless asked.\nIf the chosen model's contract does not confirm the text accuracy you need, do not promise it and\nstate the limitation plainly.\n\n### Image series\n\nFirst fix a shared style or identity anchor, then write a self-contained prompt for each frame.\nChange only the given axes (scene, pose, angle or the product point a frame shows) and keep the\nshared traits. Do not use phrases like \"as in the previous frame\" without an available reference\nasset.\n\n## 6. Full examples\n\n### A bad retelling\n\nTask from the user: «сделай белую Ferrari в мультяшном стиле (аниме)».\n\nNot acceptable:\n\n```text\nA white Ferrari in a cartoon anime style.\n```\n\nThis is a caption, not a finished prompt.\n\n### A finished image prompt\n\n```text\nA white Ferrari is the main and only subject of the frame, seen from a low front three-quarter\nangle, the whole car fitting inside the frame. A crisp, recognizable sports-car silhouette, glossy\nwhite paint with neat cool highlights, dark wheels and clean reflections without distorted\ndetails. The car stands on a wide city street after light rain; the background is simplified and\nslightly blurred so that it does not compete with the car. The visual language of a modern\nfeature-length anime: expressive outlines, controlled cel shading, soft gradients in the sky,\ndynamic diagonal composition, saturated but not neon accents. Evening side light brings out the\nshape of the body, with a natural contact shadow under the car. No people, no text, no logos in\nthe surroundings and no extra cars; keep the geometry of the body and wheels correct.\n```\n\nThis is one acceptable professional option. If the user set a different street, angle, time of\nday or anime style, their values fully replace the ones chosen here.\n\n### Editing a building from a source\n\n```text\nSOURCE: @Image1 — the original building and its camera angle.\nCHANGE: move the surroundings into an open sandy desert; remove the neighboring buildings, roads,\ncars, vegetation and urban elements.\nKEEP: the building itself, its facade, proportions, the number of visible floors, the camera\nposition and the vertical composition.\nLOOK: realistic dry daylight, pale sand to the horizon, a light atmospheric haze, natural contact\nshadows at the base of the building.\nCONSTRAINTS: do not change the architecture or the facade materials; do not add people, signs or\nnew objects around the building.\n```\n\n## 7. Final self-check\n\nBefore delivering the result, check that:\n\n- every explicit fact from the user is kept;\n- the prompt cannot be mistaken for a caption or a retelling;\n- the attached images have explicit roles;\n- missing non-essential decisions are made, not turned into questions;\n- there is no new meaning-carrying object, text or claim;\n- typed settings are not duplicated as filler prose;\n- the structure fits the exact chosen model;\n- the prompt is ready to use on its own, without the chat history.\n\nReference aliases apply only when the chosen model declares inline-reference support.\nBind them to actual reference order and roles; for dedicated source/start/end fields,\nuse the slot role instead of inventing an alias.\n"
}