Communitygithub.com

Aesthetic Director Core

A creative direction layer that turns a vague request into a single committed design idea before any blueprint or prompt is written. It runs a sequence of generative "moves" (find the obsession, plant a contradiction, pick from the tail, build causality, subtract, keep one strange thing) and produces a short Creative Brief with rejected alternatives. Use for any OC / character / illustration / fashion / key-visual request, whenever output feels generic, "AI-flavored", too plain, too busy, or when the user asks for taste, direction, aesthetics, creative direction, human touch, sophistication, or to avoid the AI feel.

What is Aesthetic Director Core?

Aesthetic Director Core is a decision-making system designed to eliminate the generic "AI flavor" from creative outputs (characters, illustrations, fashion, key visuals). It forces the agent to follow a human designer's decision path before writing any blueprint or prompt: first define a single obsession, plant a contradiction, sample from the tail of distributions, build causality chains, subtract rather than add, and deliberately preserve one strange detail. The skill runs a sequence of nine creative moves (M1–M9), generates three non-overlapping directions, explicitly rejects two with one-line reasons, and outputs a concise Creative Brief with hard rules like replacing adjectives with noun+verb+position, prioritizing subtraction over addition, and keeping exactly one "weird" element. It integrates with `personal-identity-profile` (reading taste signatures and forbidden zones) and downstream tools like `character-design-engine`. The skill includes reference files for creative moves, taste calibration pairs, anti-AI patterns, and feedback diagnosis—making it ideal for agents that need to produce high-quality, taste-driven, non-generic designs.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/Tera-Dark/personal-agent-skills/tree/HEAD/00_core/aesthetic-director-core

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

What does Aesthetic Director Core do?

0. Why This Skill Exists

The "AI flavor" is not a rendering problem—it's a decision-making problem. The model fills every slot with the most probable answer: silver hair, red eyes, black dress, standing elegantly, looking at the camera, mysterious atmosphere. Each item is "correct" in isolation, but together they are lifeless.

Human designers don't work that way. Humans first become obsessed with one thing, then let everything else make way for it; humans reject their first three ideas; humans achieve completion through subtraction rather than accumulation; humans deliberately leave one illogical element.

This skill's task is to force a human decision path before any blueprint or prompt is written. It outputs not an image, but a Creative Brief that has already made choices.

Upstream: personal-identity-profile (reads taste signature and no-go zones) Downstream: character-design-engine / illustration-direction (expands Brief into blueprint) → model adapter

1. Hard Rules

  • Decide on one core idea first, then act. If the core idea cannot be stated in one sentence, do not proceed to the next layer.
  • Must reject. Generate at least 3 directions each time, explicitly eliminate 2 with a one-sentence reason. The rejection reasons must be written for the user to see (one line each).
  • Ban adjectives in place of decisions. elegant / mysterious / intricate / ethereal / sophisticated / exquisite / mysterious are not design decisions. When they appear, replace them with noun + verb + location.
  • Subtraction before addition. When the user says "too plain," first check the proposition and contour, not add decorations.
  • Keep one strange thing. Every design must have at least one detail that "a review committee would cut," along with a reason for keeping it.
  • Don't preach. Do not output aesthetic theories to the user, do not restate this document's principles, do not give empty praise. Only give judgments and results.

2. Nine Creative Moves

Execute in order. Simple requests may combine moves, but M1 / M3 / M6 / M7 cannot be skipped. Detailed techniques and examples are in references/creative-moves.md.

#MoveOne-Sentence DescriptionWhat It Fights Against (AI Flavor)
M1Find the obsessionWhat is this design about? Noun + verb, not a job label.A bit of everything, nothing in particular
M2Plant a contradictionIdentity vs behavior / material vs occasion / polished vs broken. Only one.The character is just a stack of dimensions
M3Pick from the tailFor at least one core attribute (hairstyle, silhouette, palette, prop), write down the first 3 that come to mind and discard them all.Every attribute is the mode of the distribution
M4Build causalityPose ← Prop ← Vocation ← World. The composition must show at least two "because" relationships.Pose, clothing, and props are randomly independent
M5Choose the momentNot a state, but the 0.5 seconds before or after an action.standing, looking at viewer
M6SubtractReduce until removing one more item would break the proposition. Write down what was removed.Achieving "completeness" by adding things
M7Keep one strange thingA detail that is slightly off, slightly excessive, slightly ugly — it's the source of memorability.Everywhere safe, everywhere mediocre
M8Uneven densitySpecify one dense zone and one quiet zone. Name them explicitly.Decoration evenly spread over the whole body
M9Name the punctum≤5% accent color/accent object, placed at a specific location.Multiple high-saturation areas competing with each other

3. Execution Flow

Read personal-identity-profile (signature + no-go zones + explicit requirements for this round)
   ↓
M1–M3: Generate 3 non-overlapping directions → Eliminate 2 (with reasons)
   ↓
M4–M5: Build causality chain and define moment for the surviving direction
   ↓
M6–M9: Subtract, keep strange, density map, punctum
   ↓
Output Creative Brief
   ↓
Hand off to character-design-engine / illustration-direction

Time budget: The Brief itself should be short. The user needs judgment, not process. Internal reasoning can be long, but the output must converge.

4. Output Contract: Creative Brief

【Direction】One sentence, noun + verb. (e.g., A snake-god girl being enshrined is biting the red string of sacrifice.)
【Contradiction】One. (e.g., Ritualistic whiteness vs blood at the corner of the mouth)
【Silhouette Strategy】Choose one of five: halo/radial · vertical spear · asymmetric cascade · cocoon/volume · fragmented orbit
【Causality Chain】A because B, B because C. (At least two "because" relationships.)
【Moment】Specific description of the action 0.5 seconds before/after.
【Density Map】Dense zone: ___ / Quiet zone: ___
【Punctum】Color/object + precise location.
【Strange Thing Kept】What it is + why it is kept.
【Things Removed】List 2–4 items.
【Rejected Directions】2, each with a one-line reason.

The Brief may be in Chinese or English, following the user's current language. No explanatory paragraphs; explain only if the user asks.

5. Feedback Interpretation

User feedback is evidence about a specific layer, not permission to add decorations. See the layer-by-layer diagnosis table in references/feedback-diagnosis.md. Quick reference:

User saysFirst checkDon't do
Too plain / not memorableM1 proposition, silhouette strategyAdd accessories, add light effects
Too messy / too muchM6 M8: Remove secondary anchors, restore quiet zoneChange colors
Doesn't feel like OC / feels like a templateM2 contradiction, M3 tail sampling, motif grammarChange job label
No human touch / looks like a dummyM4 causality, M5 moment, M7 keep strangeAdd "natural expression"
Too weird / can't acceptM7 converge to one place, rest return to signatureStart over from scratch
This version is okayLock the successful dimensions, only modify the mentioned dimensionReset entire style

6. Pre-Output Self-Check

  • Direction — Can it be restated in one sentence with a verb?
  • Are the rejected directions and reasons written down?
  • Have all adjectives been replaced with noun + location + behavior?
  • What was removed? Was anything actually removed?
  • Where is that "strange thing"? Is it only one?
  • Is the dense zone vs quiet zone clear and not adjacent in a competing way?
  • Is there only one punctum with a precise location?
  • Have you avoided the no-go zones of personal-identity-profile (watches, cyber-tech, random butterflies/roses, magic circle particles, adjective-based clothing)?
  • If this version were shown to a real human designer, would they think "someone made choices" or "someone filled out a form"?

References

  • references/creative-moves.md — Detailed techniques for the nine moves, "AI would do vs Designer would do" for each
  • references/taste-calibration-pairs.md — Paired examples: generic answer vs answer with judgment for the same request
  • references/design-calibration-examples.md — User-approved reference design breakdown, essential design decisions, forbidden shortcuts
  • references/feedback-diagnosis.md — Feedback → failed layer → corrective action
  • references/anti-ai-patterns.md — List of AI flavor patterns: pattern → cause → replacement action
  • references/emotional-design.md — How emotion and narrative translate into visible visual decisions

Individual skills in this repo

This repo contains 13 individual skills — each has its own dedicated page.

Tera-Dark/personal-agent-skills

Persistent identity and taste layer for Tera-Dark's creative work. Holds the aesthetic signature (restrained base, one strange thing, a hint of danger, fashion-grade garment construction, female-oriented OC/gacha sensibility), hard dislikes (watches, cyber/mech, random butterflies-roses-particles, adjective costumes), business goals and workflow style. Does not generate prompts. Load first for any creative, design, illustration, prompt or aesthetic-judgment task; also when the user says 我的风格, 我喜欢, 按我习惯, 个人偏好.

Tera-Dark/personal-agent-skills

Entry point for all creative requests in this skill hub. Classifies intent (OC/character design, illustration, fashion, NAI5 prompt, Anima prompt, image reverse analysis, prompt review, ComfyUI/LoRA/dataset), loads identity + aesthetic direction first, then hands off to the right specialist and model adapter. Use whenever a request involves 设计, OC, 人设, 立绘, 插画, 服装, 提示词, prompt, NAI, NovelAI, Anima, 反推, 分析图片, ComfyUI, LoRA, or when it is unclear which skill should handle a creative task.

Tera-Dark/personal-agent-skills

Model adapter that compiles a finished character or illustration blueprint into Anima-ready English prompts in a disciplined Tag block + Natural Language block format. Handles Anima-specific format contract, length budgets, positive-first output, V1 faithful / V2 enhanced modes, model profiles and artifact troubleshooting. Use when the user asks for Anima prompts, Anima 提示词, or names an Anima checkpoint. Does not design — if no blueprint exists, route through aesthetic-director-core and character-design-engine / illustration-direction first.

Tera-Dark/personal-agent-skills

Turns a Creative Brief (from aesthetic-director-core) into a complete, model-agnostic character blueprint — design thesis, silhouette architecture, visual anchors, garment engineering (base / structural / signature extension / accessory system), material contrast, palette hierarchy, behavioral pose + camera, narrative residue, presentation format — then runs a subtraction pass. Use for OC, 人设, 角色设计, 服装设计, 立绘, 高定服设, 二游角色, character sheet, fashion concept. Never writes model-specific prompt syntax.

Tera-Dark/personal-agent-skills

Model adapter that compiles a finished character or illustration blueprint into a natural-language image prompt for models that are not Anima or NAI5 — Midjourney, DALL-E / GPT Image, Google Imagen / Gemini image, Flux, SDXL-style checkpoints, or an unnamed target. Keeps parameters out of the prompt unless the target officially supports them, and labels every model-specific claim with an evidence level. Use when the user names any other image model, says 通用提示词, MJ, Midjourney, DALL-E, Imagen, Flux, SD, 或没说用什么模型. Does not design — if no blueprint exists, route through aesthetic-director-core first.

Tera-Dark/personal-agent-skills

Turns a Creative Brief into an authored, model-agnostic illustration blueprint — the image thesis, captured moment, visual motif, environment-character relationship, camera and framing, scale contrast, negative space, physical light sources, density map, narrative residue, and one coherent surreal detail. Use for 插画, 氛围图, 竖屏, 半留白, 印象风, 故事感, key visual, poster, scene illustration, cinematic composition, or whenever a character needs to be placed into a designed moment rather than displayed on a plate. Never writes model-specific prompt syntax.

Tera-Dark/personal-agent-skills

Model adapter that compiles a finished character or illustration blueprint into NovelAI V5 community-format prompts — weighted artist stack, global style layer, scene base, char1/char2 blocks, source#/target#/mutual# interaction tags, and targeted negative steering with weight::tag:: syntax. Use when the user asks for NAI5, NovelAI, NAI提示词, tag prompt, Danbooru-style prompt. Does not design — if no blueprint exists, route through aesthetic-director-core and character-design-engine / illustration-direction first.

Tera-Dark/personal-agent-skills

Decomposes a reference image into transferable design language — silhouette and mass distribution, motif grammar, garment construction, material contrast, palette hierarchy, pose logic, negative space and presentation format — instead of a tag dump. Output feeds aesthetic-director-core (for original work) or a model adapter (for faithful reproduction). Use for 反推, 分析图片, 提取提示词, 还原风格, 参考这张图, reverse prompt, image analysis.

Tera-Dark/personal-agent-skills

Reviews an existing image-generation prompt by separating design intent from model syntax — what it is trying to say, which design layer is weak (thesis, silhouette, garment structure, causality, density, punctum), what is noise, and what a rewrite should change. Use for 优化提示词, 这个 prompt 哪里有问题, prompt review, 分析这段提示词, why does this prompt not work.

Tera-Dark/personal-agent-skills

Technical placeholder for ComfyUI workflow organization, node troubleshooting, metadata and LoRA/model integration. Use when the user mentions ComfyUI, workflow json, nodes, 工作流. Does not handle design or prompt creativity.

Tera-Dark/personal-agent-skills

Technical placeholder for image dataset organization, caption/tag management and training preparation. Use when the user mentions dataset, 数据集, 打标, captions, 训练集. Does not handle visual design decisions.

Tera-Dark/personal-agent-skills

Technical placeholder for LoRA training pipelines: dataset prep, caption organization, training configuration, evaluation. Use when the user mentions LoRA, 训练, fine-tune, 炼丹. Does not decide character concepts or artistic direction.

Tera-Dark/personal-agent-skills

Evaluates a generated image or compiled prompt against the original Creative Brief and blueprint, using a fixed rubric (identity preservation, outfit binding, spatial clarity, unrequested additions, V1/V2 consistency, output contract) plus a design-layer read (thesis, silhouette, causality, density, punctum, one strange thing). Diagnoses which layer failed and routes the fix to that layer with a single-variable change. Use on any feedback round: 太平淡, 太乱, 不像, 没人味, 这版可以, 评价一下, 哪里不对, compare versions.

Related Skills