ml Rules Template

# ml_rules.md — Competition Rules Template

> This file is authored by System_Architect at the start of each competition run and loaded verbatim into every Action Node's system prompt each loop. It is the dynamic layer of the two-layer prompt architecture.
>
> Fill in all `<...>` placeholders. Delete this header block before saving the live version.

---

## Competition Context

**Name:** <competition name>
**Task type:** <binary classification | multiclass | regression | ranking | ...>
**Evaluation metric:** <AUC-ROC | RMSE | MAP@K | ...> — higher is better / lower is better
**Submission deadline:** <date or "open">

---

## I/O Paths

> All paths below are relative to the competition workspace root.

| Artifact | Path |
|---|---|
| Raw train data | `<path/to/train.csv or .parquet>` |
| Raw test data | `<path/to/test.csv or .parquet>` |
| Sample submission | `<path/to/sample_submission.csv>` |
| Processed train arrays | `<path/to/processed/X_train.npy>, <path/to/processed/y_train.npy>` |
| Processed validation arrays | `<path/to/processed/X_val.npy>, <path/to/processed/y_val.npy>` |
| Processed test arrays | `<path/to/processed/X_test.npy>` |
| Trained model artifact | `<path/to/models/model.pkl or model.pt>` |
| Metric log | `<path/to/logs/metrics.txt>` |
| Submission file | `<path/to/submission.csv>` |

All nodes must read and write to exactly these paths. Do not invent new paths.

---

## Submission Format