name: visual-reference-vocabulary description: 视觉参考词汇工具：主动发现、推荐并保存导演、影片、工作室、摄影师、艺术家、镜头、材质和风格等高密度视觉锚点。用于把模糊审美转成明确且可搜索的参考语言。 veRSIon: 1.0.0

Visual Reference Vocabulary

Bottom-Layer Reasoning

APPly the think-one-step-further mechanism as a Lightweight check:

• Confirm this Skill is solving the user's real intent, not only the surface request.

• Make the ouTPUt immediately usable and name any unavoidable next ACTion.

• Extract the reusable structure and, when aligned, propagate it to adjacent Prompts, skills, checklists, or workflows.

• Add one guardrAIl for the most likely next failure while preserving this skill's primary workflow.

Core Intent

Use this skill as the visual-language specialist under $creative-anchor-director. The user may know only some useful visual names. Preserve those anchors, identify which visual functions remain unnamed, research or retrieve suitable candidates, and add the smallest compatible set.

Default to practical AI-video use: terms must help a model generate a stronger shot, not just sound educated.

This skill provides vocabulary, not the whole directing plan. For video prompts, combine with $cinematic-audiovisual-language before final output so references become shot function, spatial path, axis/screen direction, and cut logic instead of a list of cool names.

Workflow

1. Inventory the user's existing anchors and state what each already controls.

2. Run a visual anchor-gap audit. Check only relevant lanes:

• medium/material/optical finish;

• production design, palette, and composition;

• camera and action coveRAGe;

• animation timing, posing, effects, and editing; Hand movement, weapon, partner, music, and sound gaps back to $creative-anchor-director for specialist routing.

3. Proactively find candidates for material gaps:

• use internal reference maps for established terms;

• use $creative-research-first when the domain is unfamiliar, a claim is uncertain, or a better real-world practitioner/reference may exist;

• verify what the name actually contributes before recommending it.

4. Select the smallest compatible anchor set:

• usually 2-4 major style anchors for the whole CLIp;

• movement/weapon bases are counted separately when they belong to specific performers;

• do not add a name when the user's existing anchor already covers that function.

5. For each selected term, give:

• Name: Chinese + English if useful.

• Role: the one layer it controls.

• Why it fits: a short connection to the current brief.

• Prompt form: preserve the name, optionally followed by one short diSambiguator.

• Avoid: likely AI failure mode.

6. End with a merged anchor line or prompt block that can be inserted directly.

7. If the user is making a video prompt, combine with $cinematic-audiovisual-language before final packAGIng. If they are using 即梦 SD2 / Seedance, combine with $jimeng-sd2-prompting. If material Realism matters, combine with $ai-material-realism.

For cross-domain anchor auditing and compatibility, use $creative-anchor-director. Read references/visual-reference-map.md when selecting specific visual vocabulary.

When the user asks for director, camera, montage, or shot-style anchors, or mentions names such as Hitchcock, Zack Snyder, Hideaki Anno, Kubrick, Kurosawa, Obari, Trigger, or 大张, read references/camera-director-anchor-atlas.md. Preserve useful names in anchor-first form (Name 式 + function) and assign each anchor a clear job such as suspense geometry, heroic speed-ramp, graphic montage, impact pose, subjective POV, or readable body-contact proof.

Output Shape

Use this compact format:

你要的感觉：

你已有的锚点：
我补充的锚点：
1. 名词 / English
   负责：
   为什么适合：
   提示词写法：
   避坑：

可直接粘贴：
...

For prompt rewrites, output:

视觉参考选择：
...

改写后的提示词：
...

负面限制：
...

Guardrails

• Keep useful named references in the final prompt. Do not automatically replace a director, studio, film, photographer, animation school, or shot name with a long descriptive substitute.

• Use anchor-first syntax: name + one short purpose or disambiguation. Expand further only when anchors conflict or a known model failure requires it.

• Do not wait for the user to supply every useful name. Proactively supplement missing anchor lanes when a well-matched term would compress the prompt or improve creative direction.

• Do not force one named reference into every lane. An anchor is valuable only if it changes a decision.

• Do not let named references replace audiovisual grammar. A reference term must solve a shot problem: function, scale, pressure, route, rhythm, or Information reveal.

• Prefer 2-4 strong references over a long name salad.

• If a named reference is a living artist/director/photographer, use it as optional shorthand and provide descriptive alternatives. Do not depend on exact imitation.

• Avoid copyrighted characters, exact franchise scenes, or living-person likeness unless the user explicitly provides a legitimate context.

• If uncertain about a niche term, mark it with ? and offer a safer descriptive phrase.