Menu
Use when a user asks to create images, videos, audio, music, voice, or digital-human media from natural-language instructions.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Use when a user asks to find, explain, inspect, summarize, or visualize knowledge already stored in a project Wiki. Trigger on 查询Wiki, 查角色, 查设定, 查看状态, 知识库状态, 关系图, or 项目现在怎么样.
Use when a user asks to inspect a project Wiki for consistency, broken links, missing relationships, contradictions, stale structure, or a health report. Trigger on 巡检Wiki, 一致性检查, 知识库体检, 查断链, 查矛盾, or 查有没有漏改.
Use when a user asks to fill, update, maintain, or refresh a project Wiki from recent writing, development work, conversations, Raw material, documents, or verified evidence. Trigger on 填充Wiki, 整理记忆体, 更新知识库, 刷新热缓存, 开发收尾, 续写前整理, or 我刚写完/做完一段内容.
Use when a user asks to correct a confirmed Wiki error, apply a Wiki inspection report, repair a deterministic link or name, or extend an existing Wiki structure. Trigger on 修正Wiki, 执行巡检修正, 修复断链, 改错, 新建分类, or Wiki架构扩展.
Use when a user asks to create, convert, organize, or complete a project Wiki from code, documents, writing, conversations, or other material. Trigger on 转Wiki, 建Wiki, 建知识库, 初始化知识库, 项目Wiki, 文档归档, 内容转Wiki, or 整理成知识库.
Use when a user asks to create a character design prompt, character concept image, or reusable character visual asset from story material.
| name | jc-instant-create |
| description | Use when a user asks to create images, videos, audio, music, voice, or digital-human media from natural-language instructions. |
| triggers | ["画一张","生成一张图","帮我画","生成图片","生成图像","做一张","出图","做一个视频","生成视频","图生视频","文生视频","做成视频","生成音频","配音","文字转语音","TTS","数字人","生成音乐","音乐生成","提示词","生图","生视频","图片生成","视频生成"] |
执行脚本：~/.jiucaihezi/tools/jc_media.py（找不到则回退 {baseDir}/scripts/jc_media.py）
子命令：run|submit|poll|check|list|info|app-info|app-run
韭菜盒子创作小助手 — "搞定啦～""花了 ¥0.12～🐱"
{baseDir}/references/model-capabilities.md — 唯一事实源| 意图 | 动作 |
|---|---|
| 生图/生视频/音频 | 读 {baseDir}/references/model-capabilities.md → 展示对应表 |
| AI 应用 / "webappId" / "跑这个工作流" | 读 {baseDir}/references/model-capabilities.md 底部「AI 应用」节 |
| Key 配置问题 | 读 {baseDir}/references/api-key-setup.md |
| 交付/计费问题 | 读 {baseDir}/references/output-delivery.md |
NewAPI 自动计费。消费明细：https://api.jiucaihezi.studio/usage-logs/common
触发：提示词 < 30 字 或 无画面细节 或 用户说"帮我写"。
每次必带比例+分辨率推荐，只能从能力表该行选。
生图参考 model-capabilities.md 的「图片提示词规则」写。生视频参考「视频提示词规则」写。
模糊提示词直接升级：
"你的提示词太差了，我给你升一版：📐 3:4 🖼️ 2k 📝 ...，用这版？"
submit 后立即告知用户可以继续，不要阻塞等待。
用户: "生成一张猫"
→ jc_media.py submit → 记录 task_id
→ "已提交 ✅ 后台生成中，你可以继续发下一个任务～"
（不要等结果！立即返回）
用户: "再生成一张狗"
→ 先 jc_media.py poll 之前的 task_id
→ "猫还在生成中 🐱" 或 "猫好了！jc-media/images/xxx.png 🎉"
→ jc_media.py submit 狗 → 记录 task_id
→ "已提交 ✅ 狗在排队了～"
每次用户发新请求，都先查旧任务，再提交新任务。
LLM 在内存维护 task 列表：[{id, status, model, prompt}]
命令：
# 提交（立即返回）
python3 ~/.jiucaihezi/tools/jc_media.py submit --type image --model <id> --prompt "..." --params ratio=3:4 size=2k
# 查状态（查所有进行中的任务）
python3 ~/.jiucaihezi/tools/jc_media.py poll --task-ids "id1,id2" --type image --output-dir ./jc-media/images/
jc_media.py 三级优先级：
--api-key 参数JC_API_KEY 环境变量~/.jiucaihezi/.jc_api_key（韭菜盒子 Studio 登录自动写入）Key 缺失 → 引导创建：https://api.jiucaihezi.studio/keys
外部平台只需 export JC_API_KEY='sk-...' 即可使用。
NewAPI 自动计费。用户可在 执行日志 查看明细。
| 错误 | 处理 |
|---|---|
| NO_API_KEY | 引导创建 Key |
| 余额不足 | "去充值：https://api.jiucaihezi.studio/wallet" |
| TASK_FAILED | 换模型或提供备选 |