用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/LYL1015/JarvisHub --skill flova-sports-broadcast-fantasy命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
Use when the user asks JarvisHub AI Chat or a `webHero` workflow to generate a website, landing page, homepage, brand-style web design, preview-first webpage workflow, or wants section screenshots, webpage assets, and final HTML to be produced as one staged flow.
把 assets/demo 提炼成运行时可用的视觉连续性方法论:先锁资产/角色/场景/镜头语义,再做扩镜与视频。(素材生成场景:角色一致性、视觉锚点锁定、单素材生成前的锚点规范与一次只改一个变量。)
基于用户提供的设计稿图片还原真实网页。该 skill 可用于任意前端项目,但强依赖本地已启动且已授权的 JarvisHub vision 工作流;视觉模型由服务端默认配置决定,支持外部传入 vision prompt,且不可静默降级。
基于 SOC 职业分类
正在显示 SKILL.md
| name | flova-sports-broadcast-fantasy |
| description | Use when 用户上传本人照片,要生成体育转播镜头、场边嘉宾 cutaway、球场穿越幻想、世界杯倒挂金钩或直播包装风格短片。 |
用于体育直播感与体育幻想短片:基于用户参考照制作体育直播 cutaway 或体育幻想短片,同时把身份还原、直播图文、场馆氛围和运动动作拆成可验证阶段。
imageUrl / videoUrl,不能把提交态当完成态。critic sub-agent:只读取真实媒体并评审,不生成、不补素材。blocked 项,除非本轮工具列表明确暴露对应能力。Use this skill for:
Do not use it for real sports news, factual scores, or claiming the generated video is actual broadcast footage.
Follow the JarvisHub Execution Model. Use only exposed model choices and do not assume overlay/text rendering, generated commentators, or static scorebug support. When text overlays are required, prefer post/assembly if the image/video model cannot render stable text.
For identity-preserving broadcast cutaway:
If display text is missing, ask before generation. Do not invent names.
For sports fantasy:
| Mode | Use For | Core Rule |
|---|---|---|
broadcast_cutaway | seated audience/courtside/stadium TV shot | start with still screenshot and confirm likeness |
sports_fantasy_action | pulled into screen, bicycle kick, dunk, heroic play | split impossible action into controlled shots |
The still frame is the identity gate. Do not proceed to video if the person no longer resembles the reference.
Still frame should define:
Text rule:
Use one continuous take unless the user asks for montage.
Natural action sequence:
Avoid:
Commentary:
Use when the user wants a transformation or highlight move.
For impossible/complex athletic actions, do not make one overlong prompt carry the whole sequence. Use controlled shots:
Two-shot default:
Shot_01: viewer watches a match; screen liquefies/portal opens; bright transition.Shot_02: protagonist appears on the pitch and performs a highlight move.Character continuity:
Action prompt should describe sequence by camera beat:
Use English for broadcast-style prompts if the harness performs better; otherwise use Chinese. Exact display text stays exactly as supplied.
Image prompt structure:
<<<image_1>>> preserve the subject's likeness. [sport/venue/broadcast still], [subject behavior], [camera/framing], [graphic plan], [broadcast texture], no random text.
Video prompt structure:
<<<image_1>>> <<<image_2>>> preserve identity throughout. [continuous broadcast/action sequence]. [camera behavior]. [crowd/environment motion]. [audio/SFX notes]. no subtitles, no random text, no unwanted cuts.
For overlays, separate:
Check:
If one issue fails, regenerate or post-fix only that still, clip, overlay, or transition.