From trainhub
轮询训练任务状态（Kaggle kernel / SSH tmux session / Colab runtime）。单次查询，不循环。要周期性监控请组合 /loop。触发词：查训练状态、kernel status、实验跑完没、training 进度。
npx claudepluginhub shallow-dusty/claude-plugins --plugin trainhubThis skill uses the workspace's default tool permissions.
| 参数 | 必填 | 含义 |
Creates isolated Git worktrees for feature branches with prioritized directory selection, gitignore safety checks, auto project setup for Node/Python/Rust/Go, and baseline verification.
Executes implementation plans in current session by dispatching fresh subagents per independent task, with two-stage reviews: spec compliance then code quality.
Dispatches parallel agents to independently tackle 2+ tasks like separate test failures or subsystems without shared state or dependencies.
| 参数 | 必填 | 含义 |
|---|---|---|
run | 是 | run_name（与 train-submit 对齐） |
platform | 否（默认读 .trainhub.json） | kaggle / ssh / colab |
kaggle kernels status <account>/<kernel_prefix>-<run_name> 2>&1 | grep -v -i warning
状态映射：
RUNNING → 继续COMPLETE → 调 train-harvest run=<run_name>ERROR / CANCELLED → 拉 log 诊断：
kaggle kernels output <id> -p /tmp/kg-err-<run_name>
tail -100 /tmp/kg-err-<run_name>/<script_name>.log
ssh <host> 'tmux list-sessions 2>/dev/null | grep <run_name>'
ssh <host> 'tail -30 <remote_dir>/<run_name>/train.log'
training complete 或 mAP → COMPLETEError / Traceback → ERRORColab 没有服务端轮询 API。建议约定：训练脚本末尾 curl -s POST 一个 webhook（或把 results.csv 传到 Drive），train-watch 通过检查 Drive 判定完成。
本 skill 单次查询。要每 10min 自动轮询用 /loop：
/loop 10m 跑 train-watch run=L2-P4-ema-p5-kaggle，COMPLETE 时调 train-harvest
/loop 配合时的终止条件Loop 内容里指定"所有 runs 都 COMPLETE 后 CronDelete"，避免空转。