shepsci/kaggle-skill / skills/kaggle
kaggle
FlaggedUnified Kaggle skill. Use when the user explicitly mentions Kaggle, kaggle.com, a Kaggle URL, Kaggle competitions, Kaggle datasets/models/notebooks, Kaggle forums/discussions/writeups, Kaggle benchmarks, hackathons hosted on Kaggle, Kaggle badges, or Kaggle account setup. Do not use for generic ML, GPU/TPU, notebook, dataset, benchmark, or data-science tasks unless the user clearly ties them to Kaggle.
modules/competitions/README.md
View on GitHub# Competitions
Research a competition, then run it: status, data, submissions, scores.
Hackathons are in [hackathons/README.md](hackathons/README.md).
## Research (no credential for a public competition)
```bash
python3 scripts/kaggle_skill.py brief titanic
python3 scripts/kaggle_skill.py pages titanic
python3 scripts/kaggle_skill.py pages titanic --page evaluation
```
- `brief` is one screen: metric, deadline with the time left, prize, team
size, daily submission limit, whether it is a code competition, whether it
awards medals, how much data there is, and the names of its pages. With a
credential it adds whether you have entered and your rank.
- `pages` lists the pages with their sizes. `--page NAME` prints one as text
(part of the name is enough); a page longer than 12,000 characters is cut
with a note, and `--max-chars 0` prints all of it. `--all` prints every
page, `--raw` keeps the stored HTML or Markdown.
Start with `brief`. A rules page alone is 6,000 to 9,000 tokens, so read a
page only when the question needs it.
## Find competitions (credential)
```bash
python3 scripts/kaggle_skill.py competitions --days 30
python3 scripts/kaggle_skill.py competitions --mine
python3 scripts/kaggle_skill.py competitions --category featured,research --status active
python3 scripts/kaggle_skill.py details titanic --top 10
```
- `competitions` prints two lines per competition: the deadline and the
title, then the slug, category, teams, prize and metric. Community
competitions with fewer than ten teams are left out unless `--min-teams`
says otherwise. `--search TEXT`, `--limit N`. The competitions you have
entered are marked, unless `KAGGLE_SKILL_HIDE_ACCOUNT=1` is set.
- `details` prints the data files with sizes, the top of the leaderboard and
the most-voted notebooks. A lookup that fails is reported and the others
are still printed.
## Where you stand (credential)
```bash
python3 scripts/kaggle_skill.py status rsna-knee-abnormality-detection
python3 scripts/kaggle_skill.py leaderboard rsna-knee-abnormality-detection
```
- `status`: the deadline, your rank, submissions made today and left, the
submissions still being scored, your best and latest scores, and this
week's GPU and TPU hours. Facts only.
- `leaderboard`: the top rows, your row with the gap to the leader, and the
score at each medal line when the competition awards medals. Each run saves
a snapshot under `./.kaggle-skill/leaderboard/`, and the next run says what
moved. The public leaderboard is not the final one; after the deadline,
`--private` shows the one that decides the final ranks.
## Data
```bash
python3 scripts/kaggle_skill.py download titanic ./data --unzip
python3 scripts/kaggle_skill.py download titanic --file train.csv
```
Needs the Kaggle CLI, a credential, and the competition's rules accepted on
kaggle.com. The total size is read first and a download above `--max-gb`
(default 20) is refused: some competitions hold hundreds of gigabytes.
`--unzip` extracts archives and refuses any member that would land outside
the folder.
## Submit
```bash
python3 scripts/kaggle_skill.py validate titanic ./submission.csv
python3 scripts/kaggle_skill.py submit titanic ./submission.csv -m "baseline" --expect 0.77
python3 scripts/kaggle_skill.py submit titanic ./submission.csv -m "baseline" --expect 0.77 --yes
python3 scripts/kaggle_skill.py watch titanic
python3 scripts/kaggle_skill.py ledger
```
- `validate` compares a CSV with the sample submission: columns, row count,
ids, empty values, non-finite numbers. An empty value in a text column the
sample never leaves empty is a warning (WARN), since some competitions take
an empty prediction. It checks the shape, not the predictions.
- `submit` without `--yes` runs validate's checks on the file, prints what
would be sent and how many submissions are left today, and stops. With
`--yes` it submits and adds a line to
`./.kaggle-skill/ledger.jsonl`. A code competition takes
`--notebook OWNER/NAME --version N` instead of a file.
- `watch` waits until the submission is scored, prints the score, records it,
and reports the difference from `--expect`. `--ref ID` watches another of
your submissions; one that is not among your latest 100 in the competition
is shown but not recorded, since it may belong to another competition.
- `ledger` shows the record.
Submit only after the user confirms. Read
[competition-operations.md](references/competition-operations.md) first.
The ledger and the leaderboard snapshots are plain files in
`./.kaggle-skill/`, in the folder the commands are run from. They are local
records: nothing reads them but these commands. A project that should not
carry them can list `.kaggle-skill/` in its `.gitignore`.
## Simulation competitions
```bash
python3 scripts/kaggle_skill.py episodes 56699264
python3 scripts/kaggle_skill.py episodes --replay 117015562
python3 scripts/kaggle_skill.py episodes --logs 117015562 --agent 0
```
The first lists a submission's games with rewards and opponents. The other
two save a replay or your agent's log under `./downloads/episodes/`.
## Scripts
`scripts/competition_brief.py`, `competition_pages.py`, `list_competitions.py`,
`competition_details.py`, `competition_status.py`,
`competition_leaderboard.py`, `competition_download.py`,
`competition_validate.py`, `competition_submit.py`, `competition_watch.py`,
`competition_ledger.py`, `competition_episodes.py`. Each answers `--help`.
## References
- [competition-operations.md](references/competition-operations.md): before
and after a submission
- [competition-research.md](references/competition-research.md)
- [competition-overview.md](references/competition-overview.md)
- [competition-categories.md](references/competition-categories.md)
- [hackathon-endpoints.md](hackathons/references/hackathon-endpoints.md)
- [episode-endpoints.md](hackathons/references/episode-endpoints.md)
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| deprecation-and-migrationaddyosmani/agent-skillsManages deprecation and migration. Use when removing old systems, APIs, or features. Use when migrating users from one implementation to another. Use when migrating a database schema in production, such as renaming or dropping a column without downtime (expand/contract). Use when deciding whether to maintain or sunset existing code. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| performance-optimizationaddyosmani/agent-skillsOptimizes application performance across frontend, backend, queries, and databases. Use when performance requirements exist, when you suspect performance regressions, when Core Web Vitals or load times need improvement, when N+1 query patterns need fixing, or when profiling reveals bottlenecks. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |
| agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |