MLOps Industrialization

Transform prototypes into distributable Python packages

Security Scan

The skill's files and runtime instructions match its stated purpose (generating a Python package layout); there are no network calls or secret requests, but the package-creation script lacks input sanitization and has small correctness issues you should be aware of before running.

Purpose & Capability

The name/description (turn prototypes into distributable Python packages) matches the provided artifacts: a SKILL.md describing a package generator and a shell script that creates a src/ package layout. No unrelated credentials, binaries, or installs are requested.

Instruction Scope

SKILL.md instructs running ./scripts/create-package.sh which writes files into the workspace — this is expected. However the script does not sufficiently sanitize the user-supplied package name: PACKAGE_DIR is derived only by replacing '-' with '_' so values containing '..' or other path components can create directories outside the intended folder (src/../...). Also train.py contains a likely bug (calls config.model_dump() where transform_features expects a config dict) which could cause runtime errors. There are no instructions to read secrets or send data externally.

Install Mechanism

No install spec; the skill is instruction-only with a small helper script. Nothing is downloaded or extracted from external URLs.

Credentials

The skill requests no environment variables, credentials, or config paths. The code uses standard libraries and common Python packages (pandas, pydantic) but does not require secrets or network credentials.

Persistence & Privilege

The skill does not request persistent/always-on privileges (always: false) and does not attempt to modify other skills or system-wide configuration. It writes files only when its script is executed.

Assessment

This skill appears to do exactly what it says: create a three-layer Python package layout. Before running the script, review and run it in a safe/empty directory (or container) because it will create files and directories in your working tree. Avoid passing package names containing path traversal sequences (e.g., '..' or leading slashes) — the script only replaces '-' with '_' and does not sanitize other characters. Expect to install runtime dependencies yourself (pandas, pydantic, parquet engine like pyarrow) and note the small bug in application/train.py (config.model_dump() passed into transform_features) which you may need to fix. If you want additional safety, run the script in a disposable environment or inspect/modify it to validate/sanitize the PACKAGE_NAME before use.

Convert notebooks to production packages.

Features

1. Package Structure Generator 📦

Create src/ layout:

./scripts/create-package.sh my_package

Creates:

src/my_package/
├── __init__.py
├── io/          # I/O operations
├── domain/      # Pure business logic
└── application/ # Orchestration

2. Three-Layer Architecture 🏗️

Domain (Pure)

• No I/O, no side effects
• Feature transformations
• Pure functions or immutable objects

I/O (Impure)

• External interactions
• Load data, save models
• Classes for state management

Application

• Wire Domain + I/O
• Training loops, inference

Quick Start

# Create package structure
./scripts/create-package.sh my_ml_package

# Add CLI entrypoint to pyproject.toml:
# [project.scripts]
# train = "my_ml_package.application.train:main"

Key Files

Generated files:

• src/my_package/domain/features.py - Feature engineering
• src/my_package/io/data.py - Data loading/saving
• src/my_package/application/train.py - Training pipeline

Author

Converted from MLOps Coding Course

Changelog

v1.0.0 (2026-02-18)

• Initial OpenClaw conversion
• Added package generator