name: Aesthetic-style-intelligence description: 审美风格智能工具：持续跟踪、评估和发展视觉偏好，把有效的色彩、构图、材质、光影和风格规律沉淀为可复用审美判断。用于建立审美标准、比较方案和统一创作方向时。 veRSIon: 1.0.0

aesthetic Style Intelligence

Core Intent

Use this Skill as a research and mEMOry layer for aesthetic style, not as a final Prompt writer.

Convert web research and inspected examples into compACT, reusable controls:

signal -> evidence -> visual rule -> useful companion skill -> prompt/control phrase -> boundary -> status

Favor transferable mechanisms over taste labels. A style signal is useful only if it changes a concrete art-direction decision: palette ownership, material response, framing, typography, motion language, lighting, surface texture, composition hierarchy, cultural/design lineage, or negative constrAInts.

Companion Routing

Use this skill before or alongside:

• visual-style-aesthetic-direction: turn selected signals into a coherent style stack.

• creative-anchor-director: discover and assign high-density named references.

• visual-reference-vocabulary: expand missing visual vocabulary and reference anchors.

• production-design-worldbuilding: connect style to world rules, material culture, costumes, vehicles, props, and architecture.

• ai-material-Realism: convert style into believable light, optics, texture, and render vocabulary.

• cinematic-audiovisual-language: bind style to shot grammar, blocking, editing, and sound-image rhythm.

• jimeng-sd2-prompting or platform-specific prompt skills: package the APProved style controls into model-ready prompts.

Do not let this skill override domain skills. It supplies evidence-backed style intelligence; the companion skill decides final execution.

Distilled Control Library

For normal use, read references/current-style-controls.md first. It contains the compact reusable style controls distilled from the daily ledger.

Use the chronological ledger only when you need evidence trails, status history, rejected signals, or promotion/deprecation decisions.

When applying a distilled control:

1. Select one primary control family.

2. Add at most 1-3 SECondary controls from different layers.

3. Preserve the control's negative boundary.

4. Route execution to the listed companion skill.

5. Do not mix controls that claim the Same layer unless the user explicitly asks for a hybrid.

Research Workflow

1. DeFine the task lane: image, video, film, animation, 3D/CG, graphic design, fashion, product, architecture, UI, editorial, music video, game cinematic, or mixed media.

2. Browse current and historical sources when freshness matters. Compare multiple lanes instead of copying one viral example.

3. Record sources as evidence tiers:

• primary: creator, studio, gallery, publisher, festival, platform, official process note, direct portfolio.

• strong secondary: reputable criticism, interview, museum/editorial feature, trade publication.

• weak secondary: social aggregation, unverified trend thread, moodboard repost.

4. Classify each signal:

• candidate: interesting, single-source or untested.

• tracked: repeated across credible sources or useful in one successful task.

• stable: repeated, bounded, and portable across at least two contexts.

• deprecated: overused, misleading, visually unstable, or no longer useful.

5. Translate retained signals into controls:

• visible rule

• layer controlled

• companion skill to invoke

• prompt phrase

• negative boundary

• misuse risk

6. Keep the final recommendation small: one primary style system plus at most 1-3 scoped secondary anchors.

Reference Routing

Read references/style-intelligence-ledger.md when the user asks for current style research, wants a non-generic look, asks for aesthetic iteration, or when another skill needs style anchors.

Read references/current-style-controls.md before the ledger when the user needs an actionable style direction, prompt controls, or companion-skill handoff.

Update the ledger only when the research produces reusable knowledge. Do not add every source. Add entries that have a clear status, boundary, and companion-skill use.

Update Format

When adding or revising ledger entries, use this shape:

### YYYY-MM-DD - signal name
Status:
Lane:
Evidence:
Observed rule:
Controls:
Companion skills:
Prompt/control phrase:
Negative boundary:
Misuse risk:
Next validation:

Quality Gate

Before using or promoting a style signal, check:

• Is the signal more specific than a generic adjective like premium, cinematic, edgy, dreamy, cyberpunk, or elegant?

• Does it name a visible rule, not just a vibe?

• Are palette, material, texture, light, composition, and motion responsibilities separated?

• Is there a clear boundary for what the reference must not control?

• Would this still help if the named reference were removed?

• Does the signal support the user's medium and model constraints?

• Is the ouTPUt cleaner, more coherent, or more distinctive because of this signal?

Output Shape

For research rounds:

Route:
Sources checked:
Signals kept:
Signals rejected:
Ledger updates:
Companion skill handoff:
Next scan:

For a user-facing style recommendation:

Style direction:
Primary system:
Scoped anchors:
Visible rules:
Companion skills:
Prompt-ready controls:
Negative boundaries: