luma-shared

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Luma Shared Rules

Luma共享规则

Use this skill as the common foundation for all

luma-cli

skills.

Trigger this skill when the user mentions 拾光, 拾光智能体, 拾光工具, 拾光运营套装, Luma, Luma CLI, or luma-cli.

将此skill作为所有

luma-cli

skill的通用基础。

当用户提及拾光、拾光智能体、拾光工具、拾光运营套装、Luma、Luma CLI或luma-cli时触发此skill。

Before Running Tools

运行工具前

Check auth when a command needs backend access:
bash
```
luma-cli auth status
```
Discover the exact atom before using unfamiliar parameters:
bash
```
luma-cli --json tools describe <tool_id>
```

For multi-step jobs, create or select a project:

bash

luma-cli project create <project_name>
luma-cli project use <project_name>

Install local video runtime before local rendering:
bash
```
luma-cli runtime install ffmpeg
```

当命令需要后端访问时检查认证：
bash
```
luma-cli auth status
```
在使用不熟悉的参数前，先了解具体的atom：
bash
```
luma-cli --json tools describe <tool_id>
```

对于多步骤任务，创建或选择一个项目：

bash

luma-cli project create <project_name>
luma-cli project use <project_name>

在本地渲染前安装本地视频运行时：
bash
```
luma-cli runtime install ffmpeg
```

Output Rules

输出规则

Prefer explicit
```
--output
```
paths for every step.
Keep intermediate files; do not overwrite them unless the user asks.

Use project artifacts to inspect or resume work:

bash

luma-cli project artifact list
luma-cli project artifact schema

When a command supports JSON, use
```
--json
```
for agent parsing.

每个步骤优先使用明确的
```
--output
```
路径。
保留中间文件；除非用户要求，否则不要覆盖它们。

使用项目工件来检查或恢复工作：

bash

luma-cli project artifact list
luma-cli project artifact schema

当命令支持JSON时，使用
```
--json
```
以便Agent解析。

Standard Files

标准文件

Use stable names unless the user gives a project convention:

```
step0_content_research.json
```
```
step0_content_research.csv
```
```
step0_keywords.json
```
```
step1_rewrite.json
```
```
step2_tts.wav
```
```
step3_lipsync.mp4
```
```
step4_segments.json
```
```
step4_scene_units.json
```
```
step4_materials_enriched.json
```
```
step4_material_matches.json
```
```
step4_picture_in_picture_plan.json
```
```
step4_picture_in_picture.mp4
```
```
step5_subtitle.mp4
```
```
step6_cover_frame.png
```
```
step6_cover.jpg
```

除非用户指定项目约定，否则使用稳定的文件名：

```
step0_content_research.json
```
```
step0_content_research.csv
```
```
step0_keywords.json
```
```
step1_rewrite.json
```
```
step2_tts.wav
```
```
step3_lipsync.mp4
```
```
step4_segments.json
```
```
step4_scene_units.json
```
```
step4_materials_enriched.json
```
```
step4_material_matches.json
```
```
step4_picture_in_picture_plan.json
```
```
step4_picture_in_picture.mp4
```
```
step5_subtitle.mp4
```
```
step6_cover_frame.png
```
```
step6_cover.jpg
```

Boundary Rules

边界规则

CLI atoms should stay atomic. Do not hide a full production workflow in one command.
Prompt-heavy logic, material understanding, semantic matching, and protected model behavior belong on the backend.
Local logic is for stable cross-platform work: file organization, runtime cache, ffmpeg rendering, and result formatting.
Prefer friendly names for voices, avatars, and resources. Do not ask users to manually copy internal object keys.
Never print card keys, backend tokens, or private object credentials.

CLI原子应保持原子性。不要将完整的生产工作流隐藏在单个命令中。
依赖提示的逻辑、素材理解、语义匹配以及受保护的模型行为应放在后端处理。
本地逻辑用于稳定的跨平台工作：文件组织、运行时缓存、ffmpeg渲染和结果格式化。
优先为语音、虚拟形象和资源使用友好名称。不要要求用户手动复制内部对象密钥。
绝不要打印卡片密钥、后端令牌或私有对象凭证。

Failure Handling

故障处理

If a backend task returns a task id but no file, report the task id and stop at that boundary.
If PIP material matching returns zero inserts, do not force unrelated materials into the video.
If local ffmpeg is missing, run
```
luma-cli runtime install ffmpeg
```
.
If a command fails because parameters changed, inspect
```
luma-cli --json tools describe <tool_id>
```
before retrying.

如果后端任务返回任务ID但未返回文件，报告该任务ID并在此边界处停止。
如果PIP素材匹配返回零个插入项，不要强行将无关素材加入视频。
如果本地ffmpeg缺失，运行
```
luma-cli runtime install ffmpeg
```
。
如果命令因参数变更而失败，在重试前先查看
```
luma-cli --json tools describe <tool_id>
```
。