What problem does it solve?
Provides a repeatable, educational review workflow for trainee and junior ML engineers by turning PR-contained notebooks and analysis scripts into a co-located REVIEW.md with actionable feedback, scoring, and artifacts for learning.
Core Features & Use Cases
- PR and issue context gathering: fetches PR metadata, related issues, file lists, and author info to ground the review.
- Notebook conversion and artifacts: converts notebooks to script form, extracts chart images, and co-locates original notebooks plus a pyproject for reproducibility.
- Structured, educational reviews: produces a REVIEW.md with an executive summary, scoring matrix, detailed findings, prioritized fixes, and trainee-level educational notes, and posts a PT-BR PR comment linking the artifact.
- Use Case: A tech lead reviewing a junior's Jupyter analysis PR can generate an evidence-based review, request changes, and push the review artifacts alongside the code.
Quick Start
Run the trainee-review skill against a pull request reference to generate a REVIEW.md, extracted charts, converted scripts, and a PT-BR PR comment.