Kaggle
Skill3 starsChanged yesterdayMITView on GitHub
Run private Kaggle notebook jobs remotely through the official Kaggle CLI: prepare and upload notebooks, start and monitor GPU sessions, and retrieve model artifacts. Use for @kaggle or remote Kaggle training requests.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- fmind/dottheir own version, first seen Nov 15, 2026
- fmind/dotfilestheir own version, first seen Nov 15, 2026
- shepsci/kaggle-skilltheir own version, first seen Jul 3, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- shepsci/kaggle-skill★ 96same name
- fmind/dot★ 13same name
- fmind/dotfiles★ 13same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
8 rules. 0 are common, 4 are distinctive, and 4 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
4
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 8
Use this skill's kaggle launcher, which installs the official kaggle CLI into its own namespaced Linux cache via UV when absent, …
- Very specific: line 25
Create a fresh local upload directory containing exactly the .ipynb and kernel-metadata.json:
- Very specific: line 46
Add only approved Kaggle inputs.
- Distinctive: line 64
Wait for completion, inspect errors/logs, then list expected outputs:
- Very specific: line 75
Use --file-pattern when only selected artifact names are wanted.
- Distinctive: line 82
Prefer GPU. If GPU submission is rejected/unavailable, report it before choosing CPU.
- Distinctive: line 86
Ask before first remote submission/GPU use, publishing, deletion, overwriting, accepting competition rules, enabling internet, attaching …
- Distinctive: line 90
Keep notebooks private by default.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Kaggle remote notebook training
Use this skill's kaggle launcher, which installs the official kaggle CLI into
its own namespaced Linux cache via UV when absent, authenticated with the user's API token/key. Do not
use browser/CDP automation, interactive sessions, UI clicks or local training.
Setup
The user creates an API token at https://www.kaggle.com/settings/api. Authenticate
using one of the official CLI methods: ~/.agents/skills/kaggle/kaggle auth login, KAGGLE_API_TOKEN,
~/.agents/skills/kaggle/.kaggle/access_token, or legacy
~/.agents/skills/kaggle/.kaggle/kaggle.json. Never print, commit,
copy into notebooks or expose credentials. If the kaggle CLI is absent, tell the
user; do not silently install it or use another transport.
Procedure
- Inspect the complete local notebook, adapt it for noninteractive execution and
make model artifacts explicit files in
/kaggle/working. Create a fresh local upload directory containing exactly the
.ipynbandkernel-metadata.json:{ "id": "elxecutor/your-notebook-slug", "title": "Your notebook title", "code_file": "notebook.ipynb", "language": "python", "kernel_type": "notebook", "is_private": "true", "enable_gpu": "true", "enable_internet": "false", "machine_shape": "NvidiaTeslaT4", "dataset_sources": [], "competition_sources": [], "kernel_sources": [], "model_sources": [] }Add only approved Kaggle inputs. Set
enable_internetonly when the notebook genuinely needs network access. Preserve the original notebook and avoid stale execution outputs.Submit the background GPU version once:
~/.agents/skills/kaggle/kaggle kernels push -p UPLOAD_DIR --accelerator NvidiaTeslaT4This creates/updates the private notebook and runs it remotely. It is not local execution or an interactive session. Do not use
--no-runfor training.Monitor the exact handle; do not resubmit just because a command times out:
~/.agents/skills/kaggle/kaggle kernels status elxecutor/your-notebook-slug ~/.agents/skills/kaggle/kaggle kernels logs elxecutor/your-notebook-slug --follow --interval 10Wait for completion, inspect errors/logs, then list expected outputs:
~/.agents/skills/kaggle/kaggle kernels files elxecutor/your-notebook-slugDownload outputs to a new empty directory:
~/.agents/skills/kaggle/kaggle kernels output elxecutor/your-notebook-slug -p OUTPUT_DIRUse
--file-patternwhen only selected artifact names are wanted. Do not use--forceunless the user explicitly authorized overwriting. Verify expected filenames, sizes and SHA-256 locally; never deserialize models merely to test them.
Rules
- Prefer GPU. If GPU submission is rejected/unavailable, report it before choosing CPU; do not silently change the requested resource. The API requests hardware; quota, entitlement and server policy determine what is actually granted. CPU is not needed to stop a background run: Kaggle owns the job lifecycle.
- Ask before first remote submission/GPU use, publishing, deletion, overwriting, accepting competition rules, enabling internet, attaching private inputs or external uploads. A request to train described models authorizes that single private run plan after source inspection.
- Keep notebooks private by default. Use a new slug/private workspace rather than overwriting an existing notebook unless specifically approved.
- Existing direct commands support source pull (
kaggle kernels pull) and kernel deletion (kaggle kernels delete HANDLE --yes), but deletion requires explicit approval. There is no need to start/stop an interactive worker for this flow. - Report the notebook URL/handle, terminal status, requested accelerator, logs on failure and downloaded artifact paths/hashes. Never claim success from upload or queueing alone.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

