From trainhub
配置或查看项目的训练平台默认值（Kaggle/Colab/SSH 远程 GPU）。首次在项目中使用 trainhub 前运行，或要切换默认平台、更新凭据引用、调整指标口径时使用。触发词：配置训练平台、setup training、训练默认值、切平台。
npx claudepluginhub shallow-dusty/claude-plugins --plugin trainhubThis skill uses the workspace's default tool permissions.
管理项目级 `.trainhub.json`（放在项目根），声明本项目训练任务的默认平台、运行命名、指标解析规则。让后续 `train-submit` / `train-watch` / `train-harvest` 零额外参数就能跑。
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.
管理项目级 .trainhub.json（放在项目根），声明本项目训练任务的默认平台、运行命名、指标解析规则。让后续 train-submit / train-watch / train-harvest 零额外参数就能跑。
.trainhub.json 结构{
"default_platform": "kaggle",
"platforms": {
"kaggle": {
"account": "lilliareverie",
"dataset_sources": ["lilliareverie/my-dataset"],
"kernel_prefix": "project-exp",
"enable_gpu": true,
"enable_internet": true
},
"ssh": {
"host": "kindred@100.70.77.33",
"remote_dir": "/home/kindred/runs",
"conda_env": "ml"
},
"colab": {
"notebook_template": "templates/colab_train.ipynb",
"drive_folder": "MyDrive/ml-runs"
}
},
"metric": {
"primary": "mAP50",
"source": "results.csv",
"best_fn": "max"
},
"archive": {
"weights_dir": "weights/",
"filename_pattern": "{platform}-{run_name}/",
"keep": ["best.pt", "results.csv"]
}
}
mAP50 换成 accuracy / F1 等.trainhub.json 不存在 → 向用户询问必要字段（default_platform、account/host、metric.primary），写入最小配置.trainhub.json（引用 ~/.secrets/.env 对应段），只存 account name / host alias~/.kaggle/kaggle.json（kaggle CLI 标准位置）~/.ssh/config 的 Host 别名.trainhub.json 进 git（无敏感信息）