agentic-ml

Plugin agentic-ml from marketplace agentic-ml in lawwu/agentic-ml-plugin. Commands read from the repo's manifest — yesterday.

Skills (18)

• babysit-training

  Continuously monitors a machine learning training run (local, remote, or cloud) until it completes or hits a critical issue. Supports local log files, SSH targets, process names, Vertex AI Pipeline jobs, and Vertex AI Custom training jobs. Run this automatically whenever a training job is started or

• benchmark-e2e

  Benchmarks end-to-end ML execution quality across multiple modes (no-plugin/manual, plugin-driven, and AutoGluon-backed AutoML). Automatically identifies exactly one dataset scenario (hard-fraud, hard-readmission, or xhard-churn) and runs the benchmark against that single scenario. Use when asked to

• build-baseline

  Builds and evaluates non-ML baselines (majority class, mean predictor, simple rules, naive forecasting) to establish the performance floor that ML must beat. For forecasting, also runs statistical models and a zero-shot time-series foundation model (Chronos, TimesFM, Moirai) as reference baselines.

• build-review-page

  Build one-minute HTML review pages that let a person check an agent's ML work at a gate. Each page answers one question (what am I approving, what is predicted and when, is the split leak-free, does it beat the baseline, where is it wrong, what does the cutoff cost, what else was tried) and binds ev

• check-data-pipeline

  Dry-run a preprocessing pipeline on a small sample to catch shape, dtype, value range, padding, special token, label encoding, and collation issues before full training — and, for tabular pipelines (scikit-learn, LightGBM, XGBoost, pandas), preprocessing fit outside the training split, target-encodi

• check-dataset-quality

  Profile and validate a dataset before training. Works with CSV, Parquet, JSONL, HuggingFace datasets, image directories, and database tables (PostgreSQL, MySQL, BigQuery, SQLite, Snowflake). Checks for missing values, duplicates, class imbalance, split leakage, outliers, schema drift, personal data,

• check-deployment-readiness

  Checks that a model is ready to serve before promotion. Packages it with an input/output signature, re-loads it in a fresh process and compares its predictions to training-time predictions (training/serving skew), measures latency, throughput, and memory against budgets, validates inference inputs a

• check-eval

  Evaluates a trained model — a tabular model (sklearn, LightGBM, XGBoost), a predictions file, an HF checkpoint via Trainer.evaluate(), lm-evaluation-harness, or a custom eval script. Reports metrics with bootstrap confidence intervals, a paired comparison against the baseline, slice metrics, calibra

• check-failed-run

  Diagnoses failed or unstable machine learning training runs across local logs, artifact directories, process targets, experiment trackers, and Vertex AI pipelines. Collects evidence, classifies failure mode, performs root-cause analysis, and delivers a prioritized recovery plan. Run this automatical

• demonstrate-value

  Create a visual presentation demonstrating business value of a trained model, with every figure recorded in the run's showboat evidence notebook so it can be re-run and checked. Use after evaluation and explainability to synthesize findings into a stakeholder-ready demo. Invoke when asked to show bu

• explain-model

  Generates feature importance, bias audit, and model card before promotion. Invoke automatically after evaluation completes — even if the user says "looks good, let's ship it" or "ready to deploy." Deploying without interpretability review risks hidden biases, spurious correlations, and regulatory ex

• feature-engineer

  Explores data sources and engineers leakage-safe model features from files or databases. Use when asked to create features from raw tables/datasets, design joins and aggregations for a target label, improve model signal, or map business outcomes into trainable predictor sets. Invoke this automatical

• orchestrate-e2e

  Orchestrates an end-to-end ML run, from choosing the approach and defining the target, through dataset audit, baseline, experiment plan, pipeline checks, training supervision, failure recovery, evaluation with confidence intervals, interpretability, and deployment readiness, to a final promotion dec

• plan-experiment

  Designs a structured ML experiment plan with hypothesis, model candidates, hyperparameter search space, compute budget, and experiment ordering. Invoke automatically whenever a user has a defined target and wants to start training — even if they just say "let's train a model", "I want to try a few m

• recommend-new-approaches

  Recommend new research approaches, modeling ideas, and loss function modifications after a training run completes. Use this skill whenever you want to know what to try next, get modeling improvement suggestions, explore research ideas, suggest architecture changes, propose loss function modification

• review-target

  Reviews and validates machine learning target/label definitions against business outcomes before modeling. Use when asked to define or audit the prediction target, select metrics, set acceptance criteria, detect leakage risk, or choose split strategy for a new ML project. Invoke this automatically w

• select-approach

  Helps decide what kind of ML/AI fits a problem before any modeling — no ML (rules), supervised learning, forecasting (including zero-shot time-series foundation models), unsupervised learning, embedding models, prompting an LLM (with or without RAG), LLM fine-tuning, RLHF/DPO/RLVR, or agents. Use at

• train-model

  Launch and manage ML training with pre-flight checks, native early stopping, hyperparameter sweeps, experiment tracking, provenance, and checkpoint management. Includes a leakage-safe tabular training scaffold when there is no training script yet. Wraps babysit-training for monitoring and check-fail

Similar plugins