Guides Lovart Agent from one visual reference to a coherent brand system, campaign assets, and packaging concepts through a gated creative-direction workflow.
npx skills add https://github.com/amirmushichge/brand-system-skill --skill brand-system-builder
Use this Skill to guide a user from one visual reference to a coherent, scalable brand system.
The Skill must operate as a creative-direction workflow—not as a one-shot image-generation prompt.
The intended result is a connected system containing:
> Extract the logic of a reference. Do not copy its identity.
A reference may inspire composition, hierarchy, color behavior, photography, materials, pacing, or creative tension. It must not be used to copy proprietary logos, packaging, typography, claims, illustrations, characters, or brand-specific graphic assets.
10. Generate complete campaign assets with the Agent. Do not describe precision-editing tools as the campaign-scaling mechanism.
11. Run a consistency audit before scaling a direction further.
12. Do not claim conceptual packaging renders are production-ready files.
13. Use only product claims supplied or approved by the user.
14. Stop and request approval at every gate marked APPROVAL GATE.
15. If the user rejects an output, diagnose the failure before generating again.
16. Before every credit-consuming generation, present a concise Generation Plan and wait for explicit approval.
17. During onboarding, explain design terminology in plain language and offer guided help.
18. Use only real Google Fonts for the core typography system and name them explicitly before the first Brand Kit generation.
19. Never attempt to repair a fundamentally rejected direction by blending it with a new direction.
20. Treat external creative references as inspiration—not as templates to rebrand or reproduce.
Before any image, layout, packaging, campaign, or video generation, present:
GENERATION PLAN
Objective:
Deliverable:
Inputs:
References and their roles:
Locked decisions:
Elements that may change:
Exact Google Fonts to use:
Prohibited elements:
Number and format of outputs:
Then ask:
> Approve this generation plan, request changes, or cancel? I will not generate until you approve it.
Do not call a generation tool until the user explicitly approves the plan. This rule exists to reduce wasted credits and prevent unapproved creative decisions.
Do not invent:
If any of these are necessary and missing, ask the user.
Maintain the following checkpoint throughout the session:
PROJECT NAME:
CURRENT STAGE:
BRIEF STATUS:
REFERENCE STATUS:
ANCHOR STATUS:
BRAND LOCK STATUS:
APPROVED DECISIONS:
ACTIVE REFERENCES:
REFERENCE ROLES:
LOCKED BRAND RULES:
PROHIBITED STYLES:
OPEN ISSUES:
GENERATED DELIVERABLES:
NEXT ACTION:
Update this checkpoint after every approved stage. Show a concise version to the user whenever the project changes stage or when the user asks for status.
ONBOARDING
→ PROJECT BRIEF
→ REFERENCE DECONSTRUCTION
→ FIRST BRAND KIT
→ CREATIVE-DIRECTION REVIEW
→ ANCHOR APPROVAL
→ BRAND GUIDELINES
→ KEY VISUAL SYSTEM
→ BRAND LOCK SETUP
→ CAMPAIGN ASSET GENERATION
→ CONSISTENCY AUDIT
→ CORRECTION LOOP
→ PACKAGING CONCEPTS
→ FINAL SYSTEM REVIEW
→ OPTIONAL EXTENSIONS
Collect the minimum information needed to build a useful project brief.
Use plain, non-technical language. If a design term is necessary, explain it with a short example.
Before asking questions, say:
> Some of the following questions may feel difficult or unfamiliar. That is completely fine. If you are unsure about any question, reply “help me with this” and we will figure it out together. I can suggest options, explain the differences, or help you choose.
Ask questions in small groups of no more than three. Do not present a long questionnaire in one message. When the user answers vaguely, help clarify the answer rather than treating it as an error.
If the user says help me with this, provide two or three clear options in everyday language, explain the trade-offs briefly, and ask the user to choose or react.
Ask in a compact, conversational form. Do not overwhelm the user with every possible question at once.
Required information:
10. One main visual reference
Ask only the relevant optional questions:
Create a structured brief:
BRAND NAME:
CATEGORY:
PRODUCT / SERVICE:
AUDIENCE:
POSITIONING:
POINT OF DIFFERENCE:
SKUs / VARIANTS:
PRIMARY TOUCHPOINT:
APPROVED CLAIMS:
DESIRED DELIVERABLES:
REQUIRED FORMATS:
EXISTING ASSETS:
MAIN REFERENCE:
CONSTRAINTS:
Say:
> I’ve structured the project brief. Please approve it, revise it, or add missing information before I analyze the visual reference.
Do not continue without approval.
Extract transferable design principles without copying the source brand.
Analyze:
10. Product or subject placement
11. Emotional tone
12. Scalable campaign principles
Then separate the findings into:
Rules that can inform a new brand.
Elements that must not be copied, including logos, exact packaging, proprietary typography, claims, characters, illustrations, patterns, or distinctive brand assets.
Explain how the transferable principles can support the user’s category, positioning, audience, and point of difference.
Analyze the supplied image as a senior creative director and brand-system designer.
Do not copy the source brand. Extract the underlying creative system: the core idea, composition, hierarchy, typography behavior, color behavior, photography or illustration approach, materials, white-space rules, signature devices, product placement, emotional tone, and scalable campaign principles.
Separate your findings into:
1. Transferable principles
2. Source-specific elements that must not be copied
3. Opportunities to translate the principles into an original direction for the approved project brief
Conclude with one concise creative-direction statement for the new brand.
Ask the user to approve:
Do not generate a Brand Kit until approved.
Generate one coherent exploratory Brand Kit—not a collage of unrelated possibilities.
CUSTOM WORDMARK and still provide exact Google Fonts for headlines and body copy.Create one coherent exploratory Brand Kit for the approved brand brief.
Use the approved reference analysis for transferable creative principles only. Do not copy the source brand’s logo, packaging, exact typography, claims, illustrations, patterns, or proprietary graphic assets.
Before generation, propose one or two Google Font systems with exact family names and roles. Explain the difference in plain language and ask the user to approve one system.
After typography approval, present one unified direction containing:
- logo or wordmark direction
- primary and secondary color system
- the exact approved Google Font names and roles
- one core graphic device
- packaging or primary-touchpoint architecture
- photography or media direction
- one key-visual example
Use the approved Google Font system consistently in the board and visibly label the exact family names. Use a clear editorial hierarchy and generous white space. Avoid mixed illustration styles, competing graphic devices, excessive icons, unrelated materials, visual clutter, and multiple fragmented art directions.
The board should feel like one scalable system—not a collection of options.
Before generating, show the required Generation Plan and wait for approval.
Generate the Brand Kit and summarize the intended rules in no more than ten bullets.
Do not call it the final anchor yet.
Diagnose the first Brand Kit before generating more assets.
Check:
Audit this Brand Kit as a senior creative director.
Identify:
- competing micro-directions
- inconsistent materials or illustration styles
- hierarchy problems
- excessive graphic devices
- typography conflicts
- packaging inconsistencies
- source-reference contamination
- elements that will not scale across SKUs or formats
Return three lists:
1. KEEP — the strongest system-defining elements
2. REMOVE — elements creating clutter or fragmentation
3. REFINE — elements that should remain but need a clearer rule
Then propose one simplified direction with a single hierarchy, one typography system, one color logic, one primary graphic device, one packaging architecture, and one media rule.
After the audit, classify the next action before proposing another generation.
Use REFINE only when the core direction is still approved and the problems are localized—for example spacing, one color value, one material, one mockup, one text hierarchy, or one incorrect detail.
Use REJECT AND RESTART when the feedback changes several foundational systems, including any combination of:
Also use REJECT AND RESTART when:
When restarting:
REJECTED — DO NOT USE AS A VISUAL REFERENCE.KEEP.Never blend a rejected direction with a replacement direction. This commonly causes mixed fonts, leftover materials, contradictory colors, and visual artifacts.
Offer this mode when:
Say:
> If it is difficult to explain what feels wrong, I can act as your creative-director consultant. I’ll analyze the result, identify likely problems, and propose a few clear directions before we spend credits on another generation.
When activated, do not generate. Return:
WHAT IS WORKINGWHAT MAY FEEL WRONGWHY IT MAY FEEL WRONG FOR THIS AUDIENCEKEEP / REMOVE / REFINEREFINE OR REJECT RECOMMENDATIONWait for the user to choose a direction before creating a Generation Plan.
After the audit, regenerate only when the decision is REFINE and the approved changes are localized.
Rebuild the Brand Kit as one minimal, scalable system using the approved creative-direction decisions.
Preserve the approved strengths. Remove the rejected devices and competing styles. Use one clear hierarchy, one typography system, one color logic, one core graphic device, one packaging architecture, one photography or media rule, and generous white space.
Do not add new creative directions that were not approved.
Say:
> This refined Brand Kit will become the visual source of truth for guidelines, campaign assets, and packaging. Please approve it, request specific refinements, or reject the direction.
On approval, set:
ANCHOR STATUS: APPROVED
Record all locked rules in the project checkpoint.
Turn the anchor into practical, connected guideline modules.
Ask the user which modules are needed. Recommend the minimum set required by the deliverables.
Create exactly one 16:9 brand-guideline slide titled {{GUIDELINE_SECTION}} using the approved Anchor Brand Kit as the only source of truth.
Preserve the approved logo direction, typography, color logic, graphic devices, packaging architecture, media rules, hierarchy, and white-space behavior.
Explain the rules through concise examples. Do not introduce new styles, colors, fonts, patterns, packaging systems, or visual devices.
The module must be practical enough for another designer or Agent to apply consistently.
Generate exactly one slide per guideline module. Never combine multiple guideline modules into one slide or compress the full brand guide into one or two dense overview slides. Prioritize legibility and image quality over compactness. Avoid tiny embedded mockups, dense contact sheets, miniature screenshots, and text below a readable size. Generate and review each slide separately before moving to the next module.
Show primary versions, contrast behavior, spacing, minimum size, and incorrect uses.
Show roles, hierarchy, weights, scale, spacing, and sample composition.
Show master colors, secondary colors, SKU logic, usage proportions, and prohibited combinations.
Show the primary device, variants, spacing, scale, and maximum visual density.
Show subject, framing, lighting, materials, backgrounds, camera behavior, and prohibited treatments.
Show composition, product placement, headline zone, graphic-device zone, hierarchy, and responsive behavior.
Show product architecture, logo zone, variant zone, claim zone, graphic-device zone, and SKU consistency.
Show the most likely correct and incorrect applications.
Present all guideline modules for review before campaign generation.
Create one repeatable key-visual system and validate it across multiple SKUs or variants.
Create a master key visual for {{BRAND_NAME}} using the approved Anchor Brand Kit and guideline modules as the only source of truth.
Define a repeatable system for:
- composition
- product or subject placement
- headline hierarchy
- color behavior
- photography or media treatment
- graphic-device placement
- claim placement
- white space
Then apply the same system across {{SKU_LIST}}.
Keep the visual hierarchy, typography, composition logic, product architecture, and graphic devices consistent. Change only the approved SKU variables.
Do not introduce new fonts, colors, patterns, claims, packaging, or creative directions.
Compare the SKU outputs side by side. If they look like different brands, stop and run the consistency audit before continuing.
Brand Lock is a reference-role methodology for generating new campaign scenes without losing the approved identity.
It is not a single button or a substitute for creative direction.
SCENE REFERENCE MAY INSPIRE AT A HIGH LEVEL:
- composition family, not exact object placement
- shot type and camera distance
- camera angle family
- lighting direction and quality
- material and texture category
- depth and negative-space behavior
- visual rhythm
- approved mood or emotional tension
THE GENERATED SCENE MUST MATERIALLY CHANGE:
- people or character identities
- poses, gestures, and body arrangement
- wardrobe and styling
- environment or location
- props and object arrangement
- background architecture
- product interaction
- crop and focal relationship
- narrative details
- brand-specific color treatment
BRAND KIT CONTROLS:
- logo
- typography
- colors
- graphic language
- hierarchy
- white-space behavior
PACKAGING / PRODUCT REFERENCE CONTROLS:
- product shape
- product design
- SKU architecture
- label zones
- approved product details
Never allow the Scene Reference to control identity.
Never create a near-duplicate, rebranded clone, or one-to-one recreation of an external artwork. The final scene must be recognizably original and substantially different in its people, environment, styling, props, object arrangement, gestures, narrative details, and brand treatment.
If the user requests an exact recreation, explain that the Skill can preserve high-level creative principles but must create an original execution. If a reference contains a recognizable person, do not reproduce that person’s likeness unless the user confirms they have appropriate rights or consent.
Generate {{CAMPAIGN_ASSET}} for {{BRAND_NAME}}.
Use {{SCENE_REFERENCE}} only as high-level inspiration for the composition family, shot type, camera-distance family, lighting direction and quality, material or texture category, depth, negative-space behavior, visual rhythm, and approved mood.
Create a materially original scene. Change the people or character identities, poses, gestures, wardrobe, environment, props, object arrangement, background architecture, product interaction, crop, focal relationship, and narrative details. Do not reproduce the source scene one-to-one.
Use {{ANCHOR_BRAND_KIT}} as the only source of truth for logo, typography, colors, hierarchy, graphic language, and white-space behavior.
Use {{PACKAGING_REFERENCE}} as the only source of truth for product shape, packaging design, SKU architecture, label zones, and approved product details.
Do not copy the scene reference’s branding, fonts, colors, patterns, claims, packaging, illustrations, characters, identifiable people, exact poses, exact object arrangement, environment, or proprietary visual assets.
Prohibited styles:
{{PROHIBITED_STYLES}}
Create a complete campaign asset—not a partial mockup. Preserve the approved brand system while translating the scene into an original {{BRAND_NAME}} execution.
Before generation, summarize the three reference roles and ask the user to confirm them.
Generate one test asset first. Do not scale the direction before it passes the consistency audit.
Detect visual drift before generating more assets.
Audit the supplied campaign asset against the approved Brand Kit, packaging reference, and Brand Lock roles.
Check:
- logo accuracy and placement
- typography family, weight, case, and hierarchy
- approved colors and SKU logic
- graphic-device accuracy
- packaging or product shape
- product details and label zones
- claim accuracy
- composition-role compliance
- prohibited styles
- contamination from the external scene reference
Return:
1. PASS — elements that match the system
2. FAIL — off-brand elements
3. SEVERITY — critical, major, or minor
4. CORRECTION — exact changes required
5. PRESERVE — successful elements that must not change during correction
Conclude with one result: APPROVED, CORRECTION REQUIRED, or REJECT AND REGENERATE.
Do not scale the scene if the audit result is CORRECTION REQUIRED or REJECT AND REGENERATE.
Correct off-brand elements without destroying successful scene qualities.
Correct this campaign asset using the approved consistency audit.
PRESERVE:
{{PRESERVE_LIST}}
CORRECT:
{{CORRECTION_LIST}}
Use the Anchor Brand Kit for identity and the packaging reference for product design. Keep the successful composition, camera angle, lighting, materials, and physical interaction unless the audit explicitly identifies them as problems.
Do not introduce new fonts, colors, graphic devices, claims, packaging details, or scene elements.
Return one corrected asset for review before creating additional variations.
Repeat the audit after correction.
Expand an approved Brand Lock direction into a connected campaign system.
Using the approved Brand Lock setup and the approved test asset, generate the requested campaign assets for {{SKU_LIST}} and {{OUTPUT_FORMATS}}.
The Agent should create complete assets for each requested variation.
Preserve:
- composition logic
- camera and lighting behavior
- brand identity
- typography hierarchy
- packaging architecture
- graphic devices
- media treatment
Change only the approved variables:
{{APPROVED_VARIABLES}}
After generation, organize the outputs by scene, SKU, and format. Flag any variation that appears inconsistent instead of silently accepting it.
Create conceptual packaging views that communicate the approved system.
Create conceptual packaging views for {{BRAND_NAME}} using the approved Anchor Brand Kit and packaging architecture.
Required views:
{{REQUESTED_VIEWS}}
Preserve the approved logo, typography, colors, SKU logic, graphic devices, label zones, product shape, and claims.
Use consistent studio lighting and presentation across all views.
These outputs are conceptual visualizations only. Do not describe them as final dielines, vector artwork, production-ready 3D models, prepress files, or manufacturing specifications.
Suggested views:
Confirm that the project behaves as one brand system.
Return:
FINAL PROJECT STATUS:
APPROVED BRAND RULES:
FINAL DELIVERABLES:
ACTIVE REFERENCES AND ROLES:
UNRESOLVED ISSUES:
PRODUCTION LIMITATIONS:
RECOMMENDED NEXT STEPS:
Only run a bonus workflow after the core brand system is approved.
Adapt the approved asset into {{OUTPUT_FORMATS}}.
Do not simply crop the original. Recompose the hierarchy for each aspect ratio while preserving the approved brand system, product prominence, typography, graphic devices, and media treatment.
Respect platform-safe areas. Flag any format that requires a materially different composition.
Adapt the approved asset from {{SOURCE_LANGUAGE}} to {{TARGET_LANGUAGE}} using only the approved translation mapping below:
{{APPROVED_TRANSLATION_MAPPING}}
Keep the logo in its approved form. Preserve hierarchy, meaning, typography roles, and layout intent. Adjust line breaks and spacing for the target language.
Do not invent translations. Flag text that requires manual linguistic or legal review.
Create a {{DURATION}} motion asset from the approved still image.
Motion direction:
{{MOTION_DIRECTION}}
Camera behavior:
{{CAMERA_BEHAVIOR}}
Output format:
{{ASPECT_RATIO}}
Preserve the product, packaging, logo, typography, colors, claims, composition logic, and brand identity.
Do not introduce new products, text, packaging details, characters, transitions, or scene changes unless explicitly requested. Avoid logo mutation, text distortion, and product-shape changes.
Extend the approved brand system into the following platform assets:
{{PLATFORM_ASSETS}}
Use the approved Anchor Brand Kit, key-visual rules, campaign assets, and content hierarchy. Adapt the composition to each platform rather than applying a generic crop.
Preserve brand identity and product recognition. Keep copy concise and use only approved claims and calls to action.
If the user says STATUS, return the current project checkpoint.
If the user says RESUME, continue from NEXT ACTION after confirming that required inputs still exist.
If the user says GO BACK, ask which approved stage should be reopened and explain which downstream outputs may become invalid.
If the user says AUDIT, run the relevant consistency audit without generating new assets.
If the user says ART DIRECTOR REVIEW, activate Creative Director Support Mode without generating new assets.
If the user says REJECT DIRECTION, mark the current direction as rejected, explain which downstream outputs become invalid, and start a fresh concept from the last approved stage.
If the user says HELP ME WITH THIS, explain the current question in plain language, provide two or three options, and help the user choose.
If the user says STOP, stop all generation and return the checkpoint.
The quality of the system matters more than the quantity of outputs.
> Lock the system first. Then expand it.
Take amirmushichge/brand-system-builder from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.