# Refactor > Refactor Python modules for production quality Use when this capability is needed. - Skill: `tomevault-io/refactor-26` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/refactor-26` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/refactor-26/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/refactor-26 --- # Python Module Refactoring You are a world-class Python engineer. Refactor the target module following these rules strictly: ## Checklist (ALL must be satisfied) - [ ] No hardcoded values — all config via OmegaConf/Hydra YAML - [ ] Type hints on every function signature and return type - [ ] Google-style docstrings on all public functions/classes - [ ] Shape assertions at tensor function boundaries - [ ] Atomic function design (one conceptual task per function) - [ ] Prefer library calls (MONAI, einops, torch.nn.functional) over manual implementations - [ ] Logging via Python `logging` module (no print statements) - [ ] Custom exceptions for domain-specific errors - [ ] @dataclass for configuration containers - [ ] Numerical guard clauses (epsilon clamping where division occurs) ## Process 1. Read the target file completely 2. Identify all violations of the checklist 3. Propose the refactoring plan (what changes and why) 4. Implement the refactoring 5. Run: `~/.conda/envs/growth/bin/python -m py_compile ` to verify 6. Run: `~/.conda/envs/growth/bin/python -m pytest tests/ -x -q` if relevant tests exist $ARGUMENTS --- > Converted and distributed by [TomeVault](https://tomevault.io/claim/mariopasc) — claim your Tome and manage your conversions.