/meta--skills-init

Detects your project languages, frameworks and tools, then writes a commented .gitignore, merging with any existing one.

Generates or tunes tsconfig.json for a backend, frontend, library or monorepo, with strict mode and an explanation of the settings.

Generates or tunes your ESLint config for TypeScript, React or Node, with rules, ignores, lint scripts and a pre-commit hook.

Finds every environment variable your code uses and generates a .env.example, a validation schema and a config module that fails fast.

Generates or optimizes a multi-stage Dockerfile for your project, with a minimal image, layer caching and a non-root user.

Generates GitHub Actions workflows for CI, CD, auto-labeling and other automation in your repository.

Turns a raw dataset into a clean, well-split, leakage-free base for modeling: profiling, cleaning, train/validation/test splits and balance checks.

Designs features for a tabular ML problem inside a leakage-safe pipeline: encodings, scaling, interactions and date-based features.

Gives an ML project reproducible training, experiment tracking, a model registry and CI, sized to the team instead of a full platform.

Takes a trained model to production: batch or online serving pattern, prediction API, packaging, monitoring, drift detection and rollback plan.

Evaluates a model beyond one headline number: metrics matched to error costs, calibration, per-slice performance and error-by-error analysis.

Trains a model methodically: a trivial baseline first, then the simplest model that could work, with cross-validation and tuning.