提示词导演

SKILL前端设计
前端设计 未标注权限 许可证不明确
安装这个技能

运用卓越的AI摄影艺术指导与提示词(prompt)设计能力,打造连贯的多图视觉系列。在设计具体画面之前,先确立一套权威的“系列设定规范”(Series Bible),涵盖时间、地点、季节、天气、氛围、建筑、角色、服装、道具、光影、色调、摄影语言及写实度准则。后续每一张画面的提示词均须承袭这套既定设定,仅根据叙事与镜头需求进行必要的差异化调整。最终交付内容仅包含单张图像的提示词,不含内部制作规格、分析、元数据、风格锁定参数、负向约束或说明性文字。

一句话选型结论

按资料字段整理

提示词导演是前端设计类的 Agent Skill,安装方式未标注,权限级别未标注,许可证不明确,有可直接复制的安装命令。技能本质是给模型读的说明书,效果取决于模型与客户端;装完先用一个小任务验证,再放进正式流程。

分类前端设计 安装方式未标注 权限级别未标注权限 许可证许可证不明确
累计安装 未标注 本周 未标注
GitHub 星标 未标注 仓库 未标注
下载量 未标注 技能市场下载数
权限级别 未标注权限 安装前请确认风险范围
许可证 许可证不明确 未说明

安装

未标注
在支持 Skills 的客户端里执行,装完问一句「你有哪些技能」验证
skillhub install user_9ba9d3b3/photography-prompt-director
  1. 1
    确认客户端支持

    确认你的客户端支持 Skills(Claude、Codex、Cursor 等),并升级到较新的版本。

  2. 2
    拿到技能文件

    执行上面的命令安装;也可以把仓库里的 SKILL.md 手动放进客户端的 skills 目录。

  3. 3
    刷新技能列表

    重新打开客户端,或刷新技能列表,让新放进来的技能被加载。

  4. 4
    验证是否生效

    问一句「你有哪些技能」,确认列表里出现它,再拿一个小任务试跑。

技能档案

前端设计
字段来自技能仓库与公开索引,中文说明由本站标注
作者 / 维护方user_9ba9d3b3仓库 未标注 · 版本 未标注
宿主客户端未标注技能可被这些客户端加载
安装方式未标注安装包 community
权限级别未标注权限安装前请确认风险范围
许可证未说明许可证不明确,商用前请自行确认条款
仓库可用性available技能仓库当前是否可访问
最近核验未标注上游最后核验时间 · 数据更新 1791623022
正文体量13.7 KBSKILL.md 原文大小,越大说明说明书写得越细

典型使用场景

3 类
按分类整理的常见用法,具体以 SKILL.md 为准
出页面与组件

按描述生成页面结构和组件草稿。

调整样式与布局

改间距、配色与响应式布局。

生成设计说明

把设计稿整理成开发能接的说明。

选择建议

装之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有可直接复制的安装命令,不用自己翻文档找入口。

先想清楚的地方

  • 没有标注权限级别,实际能做什么要装好后确认。
  • 许可证不明确,商用或二次分发前先到仓库确认条款。
  • 没有做站内核验,内容以作者仓库为准。
  • 技能只是说明书,实际效果取决于模型与客户端,装完先用小任务试一遍。

更适合

  • 需要在前端设计场景里给模型补一套做法的人。
  • 愿意先读一遍 SKILL.md、确认它会让模型做什么的人。

不太适合

  • 对许可证要求很严、必须明确可商用的人。
  • 只想装完就自动变强、不打算看说明书的人。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
技能权限级别安装方式累计安装宿主客户端安装命令
提示词导演当前 未标注权限 未标注 未标注 未标注 有
前端设计 只读 GitHub 仓库 882.3 万 Codex、Claude、Cursor 有
Vercel 组合模式 联网访问 GitHub 仓库 385.9 万 Codex、Claude、Cursor 有
Vercel React native skills 前端 只读 GitHub 仓库 246.9 万 Codex、Claude、Cursor 有
adapt 搜索 浏览器控制 GitHub 仓库 202.7 万 Codex、Claude、Cursor 有
使用我的浏览器 浏览器控制 GitHub 仓库 184.8 万 Codex、Claude、Cursor 有

SKILL.md 全文

13.7 KB
模型实际读到的说明书原文,来自技能仓库

Photography Prompt Director v0.1

Mission

Design the photograph first; write the prompt second.

The core objective is high-end visual output: the result should feel like a deliberately produced professional photograph, not like an AI image that has been decorated with photography vocabulary.

This version deliberately preserves the strong visual-language behavior of the earlier generation while adding professional photography intelligence as a hidden decision layer. All planning documents are internal; the user-facing result is prompt-only.

Core hierarchy

Always optimize in this order:

  1. Visual quality and aesthetic impact
  2. Subject and artistic intent
  3. Believable photographic lighting and optics
  4. Composition and spatial hierarchy
  5. Material and human realism
  6. Project style consistency
  7. Technical specificity only where it improves the image

The prompt is not a technical specification dump. Internal production specifications may be highly detailed; the final prompt must be selective, coherent, visual, and model-friendly.

Critical principles

  1. Never jump directly from concept to a keyword pile.
  2. Never use “cinematic” as a substitute for cinematography.
  3. Never use “photorealistic” as a substitute for photographic realism.
  4. Treat the first visual-quality layer as primary; professional camera knowledge is an enhancement layer.
  5. Every image must have a clear visual purpose.
  6. Every image must have deliberate composition, focus, depth, light, material response, and tonal hierarchy.
  7. Technical parameters are internal decision tools. Include them in the Final Prompt only when they create a visible outcome the model benefits from understanding.
  8. Preserve natural, premium, tactile image quality rather than over-explaining photography mechanics.
  9. Favor sophisticated simplicity over visual overload.
  10. Avoid generic AI-art vocabulary unless it communicates a concrete visual outcome.
  11. Do not imitate a named photographer, living artist, or specific copyrighted image. Abstract qualities instead: documentary restraint, luxury campaign polish, museum-like tonal discipline, etc.
  12. Preserve intentional imperfection when it supports authenticity; never add fake film damage merely to look photographic.
  13. If a project is a photo series, design individual images to vary in distance, lens, depth, composition, and moment while sharing one visual identity.
  14. When the user gives a strong aesthetic direction, do not dilute it with unnecessary alternatives.
  15. If the brief is incomplete, make intelligent assumptions and proceed.

Quality target

Target the visual authority of:
- premium commercial photography
- high-end editorial photography
- strong contemporary/fine-art photography
- sophisticated architectural photography
- premium lifestyle photography

The target is not “Unsplash style.” Professional photographic libraries such as Unsplash are used only as a quality reference for clarity/detail, natural color, dynamic range, composition, art direction, concept, authenticity, and restrained post-processing. Unsplash itself does not accept AI-generated photography, so never represent generated work as Unsplash photography. citeturn107613search0turn107613search1turn107613search5

Hidden decision layer vs Final Prompt

The internal process may reason about many variables:
- focal length
- camera height
- subject distance
- aperture
- focus plane
- depth of field
- key/fill/negative fill
- exposure
- dynamic range
- material roughness
- atmospheric perspective
- optical behavior
- post-processing

But the Final Prompt must translate these into visible photographic outcomes.

Example internal decision:
> 35mm, f/5.6, subject 3m away, distant architecture, moderate-deep DOF.

Preferred prompt language:
> natural environmental focus that keeps the subject clear while preserving readable architectural depth behind them.

Do not mechanically print every parameter into the prompt.

Workflow

The defining rule of v0.4 is:

> Do not design individual pictures first. Design the shared world first, lock it, then shoot inside it.

For every multi-image request, run the following sequence internally. The Series Bible is an internal control document and must never appear in the final answer.

Phase 1 — Series brief interpretation

Extract the project's:
- subject and story premise
- total number of images, if specified
- intended audience / use
- primary photographic mode
- emotional arc
- required visual beats

If the brief is incomplete, make coherent assumptions instead of creating different assumptions for different shots.

Phase 2 — Mandatory Series Bible lock

Before designing Shot 01, create one authoritative Series Bible for the entire set.

The Series Bible must explicitly lock:

A. Time

  • year / era
  • season
  • month or date when materially relevant
  • time of day
  • temporal progression rules across the set

Do not casually move between morning, afternoon, golden hour, night, or another season. Any time transition must be intentional.

B. Place

  • country / region / city
  • exact type of location
  • recurring architecture
  • street / interior / landscape identity
  • geographic cues
  • spatial scale
  • environmental materials

A new camera angle must still read as the same location unless the narrative explicitly moves locations.

C. Weather and atmosphere

  • weather state
  • cloud cover
  • precipitation
  • humidity
  • air quality / haze
  • ground wetness or dryness
  • wind strength
  • seasonal atmospheric behavior

Weather is a continuity variable, not a shot-by-shot decoration. Do not make one image rainy and another dry without an intentional transition.

D. Environment and set dressing

Lock recurring:
- architecture
- furniture / infrastructure
- signage language
- vehicles
- props
- vegetation
- surface condition
- background population density
- recurring visual motifs

New objects may be introduced only when narratively justified and compatible with the established world.

E. Characters

For every recurring person, lock:
- approximate age
- gender presentation when relevant
- body type / silhouette
- hair style and color
- facial identity cues
- wardrobe
- footwear
- accessories
- carried objects
- baseline expression / behavior

Character identity must persist across shots. Do not redraw the same person as a different person because the camera angle changes.

F. Character state continuity

Also lock the state that can change only deliberately:
- cleanliness / wetness
- injuries or marks
- makeup / grooming
- clothing damage
- possession of props
- emotional progression
- physical position when a sequence is continuous

A recurring character cannot spontaneously become dry, clean, injured, uninjured, differently dressed, or differently styled without narrative cause.

G. Lighting and sun/weather logic

Lock:
- dominant ambient light
- sun direction when applicable
- artificial-light family
- color-temperature relationship
- shadow softness
- contrast level
- practical-light behavior
- recurring highlight behavior

The same series should not randomly alternate between contradictory lighting setups.

H. Color and tonal system

Lock:
- dominant palette
- secondary palette
- saturation level
- contrast
- black-point behavior
- highlight behavior
- skin-tone treatment
- color separation rules

Individual shots may have local color changes only when physically and narratively consistent with the locked environment.

I. Photographic language

Lock a family rather than a single identical camera setup:
- primary focal-length family
- acceptable focal-length range
- camera-height philosophy
- perspective character
- DOF philosophy
- optical character
- motion-rendering character
- post-processing character

A series can vary from wide environmental frames to tighter portraits, but it should still feel shot by the same visual system.

J. Realism and exclusion rules

Lock:
- realism level
- anatomy / material realism
- acceptable stylization
- forbidden visual drift
- forbidden genre leakage

Examples of drift to prohibit when relevant:
- cyberpunk appearing in a historical setting
- fantasy fog appearing without environmental cause
- wardrobe suddenly becoming futuristic
- neon appearing solely because a prompt asks for “cinematic” imagery
- changing architecture between angles
- random lens artifacts

Series Bible authority rule

Once locked, the Series Bible outranks every individual shot idea. If an individual shot conflicts with the Bible, modify the shot concept—not the Bible—unless the user explicitly asks for a world/time/location transition.

Phase 3 — Continuity audit before shot design

Before writing any image prompt, internally create a continuity checklist containing only immutable or slowly changing facts:

TIME / PLACE / WEATHER / ATMOSPHERE / ARCHITECTURE / RECURRING CHARACTERS / WARDROBE / KEY PROPS / LIGHTING FAMILY / COLOR SYSTEM / PHOTO LANGUAGE

Every shot must inherit these values first.

Phase 4 — Shot role planning

Assign each image one unique narrative or visual purpose, such as establishing, environmental portrait, interaction, detail, transition, action, emotional close-up, spatial reveal, or ending.

Do not reinvent the world for each shot. Variation should come primarily from camera position, framing, focal length within the locked family, subject distance, composition, gesture / action, focus hierarchy, and moment selection.

The set should feel like multiple photographs from the same produced scene, not separate AI images that merely share an aesthetic.

Phase 5 — Shot design inside the lock

For each shot, inherit the relevant continuity facts first, then decide only the shot-specific variables:
- narrative action
- exact framing
- camera position
- camera height
- camera distance
- focal length within the approved family
- focus subject
- DOF outcome
- composition
- moment / gesture
- foreground/background relationship
- local lighting interaction

Do not independently regenerate city identity, season, weather, wardrobe, character appearance, architecture language, color system, or global lighting family unless the shot is explicitly a transition.

Phase 6 — Photography intelligence

Use professional photography reasoning internally: viewpoint, lens behavior, aperture / DOF, focus plane, lighting direction, exposure, dynamic range, material response, and optical character.

Translate technical decisions into visible photographic outcomes. Do not dump camera settings into the final prompt unless a specific model materially benefits from them.

Phase 7 — Prompt construction

Build each final prompt in this order:
1. subject / action
2. locked environment and location identity
3. locked character identity and wardrobe when present
4. shot-specific composition and camera viewpoint
5. focus / depth outcome
6. locked lighting and weather behavior
7. material and atmospheric behavior
8. locked color / tonal character
9. photographic realism and the smallest necessary exclusions

Prompt inheritance rule

Do not assume the image model remembers previous prompts. Each prompt must explicitly repeat the critical locked anchors that guarantee identity and continuity, especially time of day / season, location identity, weather, recurring character description, wardrobe, key prop, lighting family, palette, and photographic character.

Prompt consistency rule

Use the same or nearly the same wording for immutable continuity anchors across all prompts. Variation should occur in the shot-specific section.

Phase 8 — Series-level final audit

Before returning the prompts, compare the complete set against the Series Bible.

Check:
- same place still recognizable
- time consistent
- weather consistent
- recurring characters remain the same person
- wardrobe stable
- props preserved
- architecture stable across angles
- lighting physically coherent
- color system stable
- photographic language feels like one production
- each image has enough variation to justify being a separate shot
- no individual prompt introduces a new genre, season, city, weather state, or character design

If a contradiction is found, fix the offending prompt before output.

Output format

The final response must contain only the image prompts.

Do not output Series Bible, internal production specification, shot metadata, analysis, explanations, Style Lock, Negative Constraints, Model Notes, camera-spec tables, continuity checklists, or any other information outside the prompts.

For a multi-image series, output one prompt per image, in shot order.

Recommended minimal format:
SHOT 01 — <prompt>
SHOT 02 — <prompt>
SHOT 03 — <prompt>

If the user requests another numbering or file-oriented format, follow it while still outputting only the prompts.

Default output philosophy

The internal reasoning can be detailed and production-grade, but it is never exposed. The final answer is the finished prompt sheet: one self-contained, continuity-locked prompt per image and nothing else.

> One world. One time. One weather state. One cast. One visual system. Many camera views.

常见问题

5 问
答案由本站依据条目字段整理
提示词导演怎么安装?

在支持 Skills 的客户端里执行:skillhub install user_9ba9d3b3/photography-prompt-director。安装方式为未标注。装完刷新技能列表或重开客户端。

提示词导演支持哪些客户端?

资料没有标注宿主客户端,但只要客户端支持 Skills(Claude、Codex、Cursor 等),一般都能加载。

提示词导演权限高吗,安全吗?

权限级别为未标注权限。安装前先看它会读取或写入哪些内容。

提示词导演收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于技能本身和所用的模型或客户端,正式使用前按作者说明确认。

提示词导演用什么许可证?

许可证标注为 未说明(许可证不明确)。商用或二次分发前请到仓库确认条款原文。

热门专题

按场景挑技能
全部专题 →

技能内容以作者仓库为准;分类、权限级别与许可证状态的中文说明为中文整理,最近核验 未标注。