Design MVR Brief

Use this skill to turn sparse brand-owner input into a design-ready creative brief. The skill must collect only the facts the AI cannot responsibly invent, then synthesize the rest through reasoning and, when needed, current web research.

Read references/interview-playbook.md before running an interview, doing MVR synthesis, or drafting the final brief.

This repository is portable across agents. For non-Codex agents, treat this file as the entry point, config/ as machine-readable rules, prompts/ as stage-specific prompt modules, references/ as the detailed playbook, and examples/ as expected behavior.

Resource Map

• config/intake-schema.json: required owner inputs, source statuses, and missing-information ending.
• config/output-examples.json: context-specific design-output examples for Objective facts questions.
• config/brief-template.schema.json: locked final creative brief structure.
• prompts/intake-audit.md: audit what is provided, infer-authorized, and missing.
• prompts/missing-info-request.md: generate missing-information requests.
• prompts/competitor-audit.md: run real competitor and reference analysis.
• prompts/visual-motif-translation.md: translate strategy into visual motifs.
• prompts/search-keywords.md: generate broad visual-reference search keywords.
• prompts/final-brief.md: write the final locked Markdown brief.

Non-Negotiable Rules

• Match the output language to the user's input language. If the user writes in Chinese, all visible output must be in Chinese, including first-level and second-level creative-brief headings; only search keyword phrases may remain in English. If the user writes in English, all visible output must be in English.
• Do not invent owner-required inputs. Objective facts, Existing assets and Preservation rules, Purpose, Subjective intent, Core value, References, and Personal-brand supplement must be user-provided before drafting unless the user explicitly says AI may infer or decide that field.
• Separate Owner-provided facts, AI inferences, and Open risks internally throughout the work.
• Before drafting any brief, run a completeness audit against the required owner inputs. A field passes only if it is user-provided or explicitly authorized for AI inference. If any required field is missing, do not draft the brief; ask only the next necessary questions in the required collection format.
• When collecting missing owner inputs, ask only for missing collection buckets, in the required order.
• When asking for missing Objective facts, include context-appropriate examples of design outputs based on what the user has already described.
• Do not ask the user to choose colors, fonts, layout styles, trends, or adjectives unless they are hard constraints.
• Convert vague preferences into testable perception claims.
• When audience behavior, competitor visuals, or public platform norms may have changed, browse/search if tools are available; otherwise label the relevant findings as unverified.
• After the completeness gate passes, produce a finished creative brief, not raw interview notes, unless the user explicitly asks for notes.

MVR Workflow

• Intake and ledger

  • Extract user input into the required collection buckets: Objective facts; Existing assets and Preservation rules; Purpose; Subjective intent; Core value; Named competitors / adjacent references / taboo references; Personal-brand supplement when applicable.
  • Build an internal ledger: Owner-provided, AI-inferred, Need to ask.

• Completeness audit

  • Check whether required owner inputs are sufficient before synthesis.
  • Mark each required input as user-provided, explicitly authorized for AI inference, or missing.
  • If any field is missing, stop and ask for the missing facts. Do not produce a creative brief, audience analysis, competitor audit, or visual motif translation yet.
  • When the audit fails, the response must not include # Creative Brief, visual motif recommendations, competitor tables, search keywords, a visible Already provided section, or a visible Still needed section.
  • Use Already provided and Still needed only as internal audit lists. The user-facing response must have exactly two visible parts: one short refusal sentence, then Please provide the missing information with the missing questions and the required final inference-authorization sentence. Do not add any other headings or summaries.
  • Do not paraphrase the collection questions. Use the wording in references/interview-playbook.md.

• Background

  • Require owner input before moving on.
  • Establish what the brand objectively is, what design outputs are needed, and what existing assets or channels must be preserved or changed.

• Purpose

  • Prefer owner input before moving on.
  • Establish how the brand creates value for its target users and what mission the brand exists to serve.
  • If missing, ask the user. Only infer purpose options when the user explicitly asks AI to infer, decide, or propose this field.

• Objective

  • Require owner input before moving on.
  • Establish the business/design problem and the desired brand visual perception.

• Core Value

  • Prefer owner input.
  • If missing, ask the user. Only infer candidate core propositions when the user explicitly asks AI to infer, decide, or propose this field.

• Audience Analysis

  • AI owns this section after the prior owner facts are known.
  • Identify the specific audience, platform context, content consumption habits, aesthetic fatigue, and signals likely to break through.
  • Avoid generic audience labels such as "young women", "urban professionals", or "Gen Z" unless backed by precise behavioral context.

• Competitive Visual Grammar Audit

  • AI owns this section after the prior owner facts are known.
  • Analyze real, specific brands, products, studios, creators, or people as direct competitors and spirit competitors. Use supplied names first; infer a labeled reference set only when none are supplied.
  • Compare core value keywords, color behavior, type logic, composition/spacing, imagery or texture, content rhythm, and owned visual territory.
  • Produce a separate anti-reference list. Do not mix "avoid" items into positive references.

• Visual Motif Translation

  • Translate abstract brand values and ideas into concrete visual language, using competitor value-to-visual relationships as evidence.
  • Cover color, typography, layout/composition, texture/photography/illustration, and each actual output use case from Background.
  • Generate concise, broad, directly searchable keywords from the visual motifs. Prefer visual-inspiration breadth over exact narrative matching.

• Creative Brief

  • Enter this step only after the completeness audit passes.
  • Deliver a single .md creative-brief document with enough context for a designer or AI design tool to start work. When filesystem tools are available, create the .md file in the current workspace, name it creative-brief-[brand-or-project-slug].md, and return its path; do not only paste the brief into chat.
  • Use the exact creative-brief template in references/interview-playbook.md: same first-level sections, same second-level sections, same order, localized to the user's input language.
  • Include assumptions only when they could materially change the direction.

Final Output Contract

Use the MVR question-frame structure unless the user requested another format. The final creative brief must be a .md document. First-level sections must be the major questions/modules; second-level sections must be their corresponding subquestions. Use the English heading set for English input and the Chinese heading set for Chinese input.

English heading set:

• Background
• Purpose
• Objective
• Core Value
• Audience Analysis
• Competitive Visual Grammar Audit
• Visual Motif Translation

Chinese heading set:

• 背景
• 存在目的
• 目标
• 核心价值
• 受众分析
• 竞品视觉语法审计
• 视觉母题转译

Do not add a standalone section that merely repeats user-provided facts. If assumptions or risks are material, place them under the relevant major question instead of creating a separate first-level section unless the user explicitly asks for one.
Do not rename, reorder, collapse, or omit template sections. Do not output a chat-style brief instead of a Markdown document.

Quality Bar

• Every recommendation must trace back to owner facts, cited public evidence, or AI inference explicitly authorized by the user.
• Never output a creative brief when any required collection bucket has not passed the completeness audit.
• Every visual motif must be specific enough to guide design decisions.
• Search keywords must be derived from visual motifs, mostly 2-4 words, broad enough for visual discovery, and not overloaded with narrative specificity.
• Every use case must get its own keyword logic; do not reuse the same keyword set for website, social covers, newsletter banners, decks, posters, UI, or packaging.
• Remove decorative language that cannot change a designer's next decision.