菜单
Generate balanced Markdown episode_index assignment tables for LeRobot QC datasets. Use when the user asks to allocate a dataset's episodes among annotators, recreate a known QC assignment table, or add users to a future allocation.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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/Spphire/LerobotQualityCheck --skill generate-qc-assignment-table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。
Validate local or remote LeRobot datasets end to end, including v2/v2.1 metadata, episode numbering and UUIDs, Parquet readability and frame indexes, video presence and full decoding, and optional latent, instruction-embedding, and sidecar caches. Use when Codex is asked to check dataset integrity, completeness, corruption, episode/video counts, training readiness, or diagnose missing LeRobot files on an SSH server.
Develop, test, inspect, or synchronize the LeRobot Quality Check Platform without interrupting production. Use when changing platform code, using the dev worktree, testing port 18081, inspecting the dev service, or preparing a production deployment.
Generate a new LeRobot dataset from QC annotation results and place it on a training server. Use when the user wants to filter a source LeRobot dataset by accepted/rejected QC labels, auto-locate the matching annotation store for the dataset, reindex/copy accepted episodes, validate the output, or transfer/stage the filtered dataset to a specified remote host and path.
基于 SOC 职业分类
正在显示 SKILL.md
| name | generate-qc-assignment-table |
| description | Generate balanced Markdown episode_index assignment tables for LeRobot QC datasets. Use when the user asks to allocate a dataset's episodes among annotators, recreate a known QC assignment table, or add users to a future allocation. |
Unless the user asks for explanation, output only:
<dataset path> <dataset display name>.用户名, 自选 UID, and episode_index 范围.Use zero-based, inclusive, contiguous ranges. Cover every index exactly once and preserve the requested user order. Leave the UID blank when it is unknown; never invent one.
For /mnt/nm_dataset/dataset/giftbox_0628_1912episodes titled 【礼物装盒新SOP】, use:
/mnt/nm_dataset/dataset/giftbox_0628_1912episodes 【礼物装盒新SOP】
| 用户名 | 自选 UID | episode_index 范围 |
|---|---|---|
| chenwendi | wendi | 0-191 |
| xiaxinyuan | xxy | 192-383 |
| majianhan | 无敌暴龙战神 | 384-574 |
| yuwenye | Virlus | 575-765 |
| zhangdi | chuan | 766-956 |
| lvjunlin | junlin | 957-1147 |
| zouyanwen | 永雏塔菲 | 1148-1338 |
| fangjunjie | 菲比啾比 | 1339-1529 |
| yingzixi | 东雪莲 | 1530-1720 |
| chenyiming | yiming | 1721-1911 |
Use this order only when the user says to use the latest/default group and supplies no different list:
| 用户名 | 自选 UID |
|---|---|
| chenwendi | wendi |
| xiaxinyuan | xxy |
| majianhan | 无敌暴龙战神 |
| yuwenye | Virlus |
| zhangdi | chuan |
| lvjunlin | junlin |
| zouyanwen | 永雏塔菲 |
| fangjunjie | 菲比啾比 |
| yingzixi | 东雪莲 |
| chenyiming | yiming |
| xieyichen | |
| zhaotianhao | |
| yangchaoyun |
For total episodes and n users:
base = total // n
remainder = total % n
Give the first remainder users base + 1 episodes and every remaining user base episodes.
For example, 1912 episodes across ten users yields two ranges of 192 and eight ranges of 191.