luma-shared

Original：🇺🇸 English

Translated

Use before any Luma / 拾光 / 拾光智能体 / 拾光工具 production workflow. Defines common luma-cli rules for auth, tool discovery, projects, artifacts, runtime resources, and safe agent behavior.

6installs

Sourcezl007700/luma-cli

Added on2026-06-03

NPX Install

npx skill4agent add zl007700/luma-cli luma-shared

SKILL.md Content

View Translation Comparison →

Luma Shared Rules

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.

Before Running Tools

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
```

Output Rules

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.

Standard Files

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
```

Boundary Rules

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.

Failure Handling

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.