kaggle-cli - SKILL.md Agent Skill

name: kaggle-cli description: > Use the local Kaggle CLI skill for command guidance, workflows, and troubleshooting across competitions, datasets, kernels/notebooks, models, model variations and versions, inbox file uploads, forums/discussions, benchmarks, configuration, OAuth/API-token authentication, and accelerator quota. Activate this skill when the user asks about kaggle CLI commands, examples, flags, metadata files, download/upload flows, submissions, benchmark tasks, or Kaggle CLI behavior.

Kaggle CLI

Use this skill to answer or operate on the kaggle command-line tool. Treat this skill and its references as the available command guide.

Quick Start

pip install kaggle
kaggle --help

Authentication options:

kaggle auth login
# or set KAGGLE_API_TOKEN
# or place an access token in ~/.kaggle/access_token
# or use legacy ~/.kaggle/kaggle.json credentials

Command Tree

kaggle
├── competitions | c
│   ├── list, files, download, submit, submissions, leaderboard
│   ├── team-submissions, episodes, replay, logs, pages
│   └── topics {list, show}, topic-messages
├── datasets | d
│   ├── list, files, download, init, create, version
│   ├── metadata, status, delete
│   └── topics {list, show}
├── kernels | k
│   └── list, files, init, push|update, pull|get, output, status, logs, delete
├── models | m
│   ├── list, init, create, get, update, delete
│   ├── topics {list, show}
│   └── variations | instances | v | i
│       ├── get, init, create, files, list, update, delete
│       └── versions | v {list, create, download, files, delete}
├── files {upload}
├── forums | f {list, topics {list, show}}
├── benchmarks | b
│   ├── auth, init
│   ├── tasks | t {push, run, list, status, download, log|logs, models, delete, publish}
│   └── topics {list, show}
├── config {view, set, unset}
├── auth {login, print-access-token, revoke}
└── quota

Note: the CLI accepts aliases such as kernels get for kernels pull and kernels update for kernels push. Do not recommend models variations versions init; use models variations init for variation metadata instead.

Reference Map

Read only the reference needed for the user's task:

Competitions - competition discovery, files, downloads, submissions, leaderboards, simulations, pages, topics.
Datasets - dataset search, files, downloads, metadata, create/version/status/delete, topics.
Kernels - notebook/script discovery, metadata, push/pull, outputs, status, logs, delete.
Models - model records, metadata, create/get/update/delete, model topics.
Model Variations - create and manage framework-specific model variations.
Model Variation Versions - create, list, download, inspect, and delete variation versions.
Files - inbox uploads, resumable uploads, directory compression behavior.
Forums - global discussion forums, topics, and comments.
Benchmarks - benchmark auth/init, task push/run/status/download/log/model flows, benchmark topics.
Configuration - config file, default path, proxy, default competition.
Authentication - OAuth login, access token printing, revocation, token/key sources.
Quota - weekly GPU/TPU accelerator quota.

Operating Guidance

Read only the reference needed for the user's task.
Prefer kaggle <group> --help or kaggle <group> <command> --help when a flag is uncertain in the installed CLI.
For metadata files, prefer the relevant init command to generate a starter file before editing it.
Do not invent commands that are not listed in this skill. If live --help output differs, report it as a version-specific difference.