leakage-detection commands

StamKavid/last-ds-mile

Plugin Claude Code

A guided data-science lifecycle for Claude Code, with leakage and honesty checks built into every stage.

StamKavid/last-ds-mile

Plugin Claude Code

A guided data-science lifecycle for Claude Code: frame, explore, baseline, validate, model, evaluate, and report — with leakage and honesty checks built into every stage.

StamKavid/last-ds-mile

Instructions file

Instructions for StamKavid/last-ds-mile, covering claude.md, what this repo is, hard rules, working in this repo and style.

StamKavid/last-ds-mile

Agent

Fast structural profiling sweep for a dataset — shape, dtypes, missingness, cardinality, duplicate keys. Use during /ds-data or /ds-explore for a quick first-pass profile. Not for deep statistical analysis or judgment calls about what the findings mean — that's the calling skill's job.

StamKavid/last-ds-mile

Agent

Runs the ds-method discipline checklist against a notebook or pipeline before /ds-report — baseline present, validation strategy sound, slice performance checked, metric matches the problem. Use before final reporting/handoff, or when asked to sanity-check a DS pipeline end to end. Not for hunting leakage specifically…

StamKavid/last-ds-mile

Agent

Adversarially hunts for target leakage across a feature pipeline — features that encode the target directly, temporal leakage where future information reaches training data, and validation-split leakage. Use before /ds-model or /ds-report when a metric looks implausibly good, or as a final check before a pipeline…

StamKavid/last-ds-mile

Command

Rewrite the results as a one-page brief with no jargon.

StamKavid/last-ds-mile

Command

Capture a failure-and-fix pair as a project lesson.

StamKavid/last-ds-mile

Command

Show the Last DS Mile pipeline map and which stage comes next.

StamKavid/last-ds-mile

Hook

Runs when a session starts on startup, resume and clear, executing ds-python.sh via bash. From StamKavid/last-ds-mile.

StamKavid/last-ds-mile

Hook

Runs after a tool call finishes for Read, Edit, Write, MultiEdit and NotebookEdit tool calls, executing ds-python.sh via bash (2 commands). From StamKavid/last-ds-mile.

StamKavid/last-ds-mile

Hook

Runs before the context is compacted, executing ds-python.sh via bash. From StamKavid/last-ds-mile.

StamKavid/last-ds-mile

Skill Claude CodeCodex

Holds the shared discipline every stage of this pipeline cites — the Red Flags, the Common Rationalizations, and the Hard Gates. Use when someone pushes back on a gate, asks why a baseline is needed, or wants to peek at the test set just once. Use when deciding whether a stage should stop and ask or produce the…

StamKavid/last-ds-mile

Skill Claude CodeCodex

Chooses a leakage-safe way to split data — temporal, grouped, stratified, or nested cross-validation — matched to how the rows are actually structured, and decided before any model is trained. Use when someone asks how to split into train and test, set up cross-validation, or pick between KFold, GroupKFold…

StamKavid/last-ds-mile

Skill Claude CodeCodex

Handles targets where the interesting class is rare — resampling, class weights, threshold tuning, and the metric consequences. Use when only a tiny fraction of rows are positive, or when a model never predicts the rare class at all. Use when someone mentions SMOTE, oversampling, undersampling, or classweight. Use…

alex-jb/shadow-perception-mcp

Skill Claude CodeCodex

Pre-submit council for Kaggle notebooks. Five voices review the notebook for train/test leakage, protected-class proxy features, seed pinning, competition rule compliance, and submission artifact presence. Returns SHIP / REWORK / BLOCK verdict with per-voice rationale and Ed25519 attestation. Use before hitting Submit…