Design & Motion · leonxlnx/taste-skill

brandkit

Name: brandkit
Author: leonxlnx/taste-skill

Premium brand-kit image generation skill for creating high-end brand-guidelines boards, logo systems, identity decks, and visual-world presentations. Trained for minimalist, cinematic, editorial, dark-tech, luxury, cultural, security, gaming, developer-tool, and consumer-app brand systems. Optimized for intentional logo concepting, refined composition, sparse typography, strong symbolic meaning, premium mockups, art-directed imagery, and flexible grid layouts.

L1 Basic · Source registration onlyV1 Commit-PinnedDual-pass Audit

Control levels describe skill capabilities and suggested control measures; they do not represent quality or safety certifications, nor do they dictate usability. Usage decisions rest entirely with users and administrators. Different agent hosts apply varying constraints; do not assume hosts provide automatic protection. Third-party skill copyrights belong to their original authors. Use at your own risk.

1How It Works & Mechanism

Essentially an art-direction instruction set injected into image generation models (ChatGPT Images / Codex image mode, etc.): writing zero code and calling zero APIs, it alters image generation distributions purely through a 798-line pipeline of 'brand strategy first → visual system second → final image prompt last'.

skills/brandkit/SKILL.md

Create a complete brand world in one image.

This is the sole technical interface between this skill and text-to-image tools: the PROMPT TEMPLATE at the end of SKILL.md is the prompt template fed to the image model, with no scripts or network endpoints in the chain.

Reframes 'image generation' as 'brand justification first': requires each board to answer 5 questions (what the brand represents / core metaphor / how the logo expresses it / how the system extends / why it is distinct), preempting random collages with strategy-first discipline.

skills/brandkit/SKILL.md

Every generated board must answer:

What is happening here: attributes the most common failure of AI-generated images (visually pleasing yet meaningless) to 'not selecting meaning first', thus mandating an upfront brand strategy paragraph in the prompt—the first section of PROMPT TEMPLATE consists of the five Brand strategy pillars.

Symbol selection is tabulated: categorizes core concepts and selectable symbol logics by domain (developer tools/AI assistants/security/gaming/voice/compliance/drones/luxury/productivity), strictly prohibiting arbitrary symbol choices.

skills/brandkit/SKILL.md

| Developer tool | building, speed, precision, control | cursor, frame, bolt, scaffold, grid |

Translating 'logos must convey meaning' into a lookup table operation is the pivotal step enabling non-designers to produce qualified brand boards; the symbol logic across the table's 9 domains maps directly to the actual Visual Modes.

Supplies logo design methodologies rather than adjectives: 5 conceptual approaches (lettermark + metaphor, product motion, metaphor fusion, negative space, constructive geometry), capping combinations at a maximum of two to prevent overdesign.

skills/brandkit/SKILL.md

Use one or combine two maximum.

Paired with the 'do / avoid' list in §LOGO GENERATION STANDARD (banning random animals, faux-luxury crests, and copying well-known marks), transforming aesthetic standards into actionable methodological choices.

Composition is structured as a 'default 3×3 nine-panel grid with fixed panel roles', while offering alternative layouts such as 2×3 / 2×2 / 1×3 / 4×2, requiring rhythmic contrast rather than having every panel compete for attention.

skills/brandkit/SKILL.md

- Default layout: `3 × 3`

Nine-panel roles are detailed item-by-item in §DEFAULT 3 × 3 PANEL SYSTEM (Logo Cover / Logo Construction / Digital Application / Brand Essence / Color System / Typography / Physical Application / Image Direction / System Detail), modularizing the 'brand board' into a component manifest.

Presets visual modes for 11 brand archetypes (Dark Dev / Dark Ops / Dark Nature / Dark Security / Light Editorial / Luxury Beauty / Voice / Cultural Experimental, etc.), offering a quartet of use cases, visual cues, logo logic, and emotional tone for each.

skills/brandkit/SKILL.md

Choose based on the brand.

This tier acts as a 'style library': models need not invent aesthetics on the fly, but simply select a mode and populate the template; this also explains the origin of 'Trained for minimalist, cinematic, editorial, dark-tech, luxury…' in the description.

Provides a ready-to-paste PROMPT TEMPLATE and final inspection criteria, consolidating the entire document into an executable prompt plus a one-sentence acceptance rubric.

skills/brandkit/SKILL.md

Create a premium brand-kit overview image for "[BRAND NAME]".

The template contains eight sections: Brand strategy / Layout / Panels / Visual mode / Palette / Style / Typography / Logo. §FINAL OUTPUT STANDARD at the end provides the acceptance checklist (clean / strategic / symbolic / minimal / consistent / premium / art-directed / viable).

Clear boundaries for reference images: extracts rhythm and quality only (grids, spacing, type scale, density, logo placement, image treatment), strictly banning the copying of specific logos, brand names, compositions, or slogans.

skills/brandkit/SKILL.md

Use references as quality training, not as templates.

This doubles as both a copyright boundary and an anti-plagiarism statement, aligning with 'no copied real-world logos' in the PROMPT TEMPLATE.

2Core Capabilities

01Brand strategy first: deducing metaphors from category, audience, emotional promise, and cultural tier before translating into symbols

02Logo conceptual methodology (5 techniques, max 2 combined) along with logo quality benchmarks and forbidden lists

03Default 3×3 nine-panel brand board layout + 4 alternative layouts (2×3 / 2×2 / 1×3 / 4×2)

04Visual mode library across 11 brand archetypes (Dark Dev / Security / Luxury / Voice / Culture, etc.), each equipped with visual cues and emotional positioning

05Compositional pacing control: alternating quiet, functional, emotional, technical, atmospheric, and detail panels; forbidding equal loudness across all frames

06Typography and copywriting discipline: minimal text, punchy micro-slogans, strictly prohibiting lorem ipsum and generic marketing fluff

07Mockup and image direction specs: restrained applications like browser frames/terminals/badges/cards; imagery requires deliberate art direction over stock aesthetics

08Anti-generic checklist: bans floating random icons, meaningless color blocks, fake micro-UIs, cheap neon, and corporate slide-deck aesthetics

4Recommended Governance & Evidence

Suggested Controls for Safe Operation

For Agent Runtime

Can be auto-loaded

For Administrators

Evidence & Triggers (File Paths & Excerpts)

SKILL.md: Pure prompt directives and rule text with zero script execution or external impact

Audit Notes: Pure directive-based skill: contains only a single markdown file in the directory, executing zero scripts, initiating no network egress, reading no credentials, and writing no local files. Its output is an image rendered via the user's own text-to-image tool, without touching external systems. The primary user caution lies not in the skill's behavior, but in output usage: brand boards may generate logos resembling existing entities. Although built-in rules forbid copying well-known marks and faux-luxury crests, human trademark clearance remains mandatory prior to commercial deployment.

5Second-Pass Verification

[ok] 外部依赖与网络调用清单为空 — 全文 URL/端点/包管理/凭证 token 扫描零命中。与 imagegen-* 两个 image skill 一致：它们都只描述图像内容，不指定任何外部服务或包。
[ok] 『Image-generation-only / does not write code』声明的出处 — 该表述出现在 README.md Skills 表与 skills/llms.txt（'Image-generation-only skill for creating premium brand-kit overview images with logo concepts, identity systems, color palettes, typography, and mockups. Does not write code.'），SKILL.md 正文未复述此句；但正文 798 行通篇为图像创作指令，grep 'code' 零命中，声明与实际一致，非夸大。
[ok] description 里 'Trained for …' 是否暗示模型微调（可能夸大） — description 的 'Trained for minimalist, cinematic, editorial, dark-tech, luxury, cultural, security, gaming, developer-tool, and consumer-app brand systems' 对应正文 §VISUAL MODES 的 11 个模式段落（暗色开发/暗色运营/暗色自然/暗色安全/浅色编辑/奢侈美妆/语音/文化实验等），是提示词层的风格库而非真的训练，用词偏营销但可落到正文证据。
[discrepancy] 与同仓库反紫蓝禁令是否冲突 — 同仓库 stitch-skill 与 taste-skill-v1 均把紫/蓝霓虹列为全局禁令（THE LILA BAN / AI Purple）；brandkit 的 Voice / Communication 模式却写明 '- lilac glow'，另在 COLOR DISCIPLINE 留了 '- no generic purple-blue AI glow unless appropriate' 的例外口。差异来自定位不同（按品类适配 vs 通用禁令），非文件错误，但跨 skill 使用时审美口径不一致，已记录。
[ok] 参考图使用是否有版权边界 — §REFERENCE USAGE 分列 'Extract'（节奏/网格/间距/字阶/密度等抽象特征）与 'Do not copy'（exact logo / brand name / composition / slogan / unique visual asset），PROMPT TEMPLATE 亦要求 'no copied real-world logos'，边界明确。
[ok] 元数据（license / stars / last_push / 路径 / commit） — gh api：MIT、87309 stars、pushed_at 2026-08-24T15:23:56Z；git ls-files 确认 skills/brandkit/SKILL.md；HEAD == origin/main == ccbc15639c97057cbfcf32ecebc38ef716e4bb37。

6Verdict & Scenarios

A strategy-first workflow solves the 'visually pleasing yet meaningless' pitfall of AI generation upfront by establishing metaphor and symbol logic prior to visual composition.

Symbol selection is structured into lookup tables, enabling non-designers to produce well-grounded logo directions (9 categories × core concepts × symbol logic).

Layouts and contents are strictly structured: the default 3×3 nine-panel grid assigns discrete roles, complemented by 4 alternate layouts, fixing the information hierarchy before generation.

The 11 visual modes provide an instantly swappable style library, spanning diverse categories such as dark-tech, luxury beauty, security compliance, and cultural experiments.

Delivers a complete, ready-to-paste prompt template and acceptance checklist, consolidating an extensive manual into a single workflow action.

Best for: Designers, founders, and marketing teams needing rapid production of brand guideline boards, logo directions, identity decks, and visual world concepts; early-stage brand exploratory sprints (multi-direction prototyping at far lower costs than hiring an agency); users aiming to inject professional art direction into text-to-image models (ChatGPT Images / Codex image mode, etc.); indie hackers needing a single-page brand system overview for clients.

Not for: Tasks requiring production-ready code deliverables (app UIs / websites)—it only generates images; formal logo production needing editable vector sources or trademark registration files (generates conceptual boards, not editable vector assets); scenarios requiring photorealistic product photography or actual physical packaging proofs; tasks requiring strictly deterministic pixel-level reproducible outputs (text-to-image involves inherent randomness; this skill does not provide seed pinning); commercial projects intending to deploy imagery directly without trademark searches.

Install Agent-Ready

① Verified Mirror 2026-09-15

Option A · Direct DownloadDownload brandkit.tar.gz

sha256: 7f9a757fefa72b05…

Option B · Copy Guide for Agent

install.json

Agents automatically download and verify mirrors from this JSON guide.

② Upstream GitHub

GitHub Repo ↗

Pinned to commit ccbc15639c.