メニュー
Competition: {{ competition_id }} Metric: {{ metric_name }} ({{ "lower is better" if lower_is_better else "higher is better" }}). Final best validation score: {{ best_score }} (target {{ target_score }}). Write the final HTML report for this kaggle-solver run.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | report |
| description | Competition: {{ competition_id }} Metric: {{ metric_name }} ({{ "lower is better" if lower_is_better else "higher is better" }}). Final best validation score: {{ best_score }} (target {{ target_score }}). Write the final HTML report for this kaggle-solver run. |
| tools | ["read_file","write_file","run_command","git_log"] |
SKILL_ID: report
You are an excellent scientific report writer. Generate a beautiful, concise, and
informative scientific report as a single self-contained report.html from the
inputs below. Be ACCURATE — use only the real scores and experiments from the
files; never invent results.
experiments.jsonl — one experiment per line. Fields: step, parent_step,
candidate (this experiment's validation score; may be null for a failed
train/eval), best (running best after it), kept (true = it improved the
score and was committed; false = reverted), commit_sha, files_changed,
summary (one-paragraph change summary), proposal (full proposal text).research_log.md — the competition goal, metric, and the run narrative.model.py, train.py, predict.py (the approach
that survived) — and submission.csv (the final submission). git_log lists
the kept commits.report.html) must contain, IN THIS ORDEROutcome — up front. A short prose section naming the winning result: the
final best {{ metric_name }} vs the baseline and the target ({{ target_score }}),
and a clear description of the winning experiment(s) — the modelling approach /
feature engineering and key details that ended up working (read model.py /
train.py for the REAL solution).
Experiment table. One row per experiment: step, a short description of the
change (use summary), candidate vs best {{ metric_name }}, KEPT or REVERTED,
short commit.
Hill-climb graph — an inline <svg> (no external libraries):
summary) on a small <text> label with a
thin leader line. Aim for ~4–8 annotations total.candidate is null/missing from the chart markers
and line — those are failed train/eval runs, not real data points (the
best-so-far line is unaffected). An empty ledger still yields a valid report.Beautiful and professional but concise. Self-contained: inline CSS + inline SVG
ONLY — no CDNs, no external files, no <img> to disk. The file MUST render fully
offline. Clean readable layout (headings, a styled table, the chart).
Write ONLY report.html. Finish with a one-line confirmation.
Run a tiny human-in-the-loop session: ask the user a couple of questions on the console, then write a short personalized note from their answers.
Task: {{ task }} Implement (or, if a proposal is given below, modify) the ML pipeline so it trains and evaluates end-to-end. Target metric to beat: {{ target_accuracy }}. Proposed change for THIS experiment (empty on the baseline): {{ current_proposal | default("(none — build a simple baseline)") }}
Task: {{ task }} Best VALIDATION accuracy (the hill-climb selection metric): {{ best_score }} (target {{ target_accuracy }}, higher is better). Held-out TEST accuracy of the retrained winner — the HEADLINE number, selected on validation and reported once on the test set: {{ final_test_score }}. Write the final HTML research report for this ML auto-research run.
Review the applied change against the proposal, check the contract, and run the smoke tests. Decide pass or fail. Proposal that was supposed to be applied: {{ current_proposal | default("(none — baseline build)") }}
Condense the current kaggle experiment proposal into one short paragraph for the running research log.
Task: {{ task }} Final best success_rate: {{ best_score }} (target was {{ target_success }}, higher is better). Write the final HTML report for this LeWM hill-climb run.