What kind of files does this skill protect against?

It prevents the staging of secrets (.env, keys), competition data (public/private/raw), large run artifacts, and database files like SQLite.

How does it handle commit messages?

It generates a structured message using a specific Agent ID and categorizes the work into areas such as workflow, docs, or orchestrator for better log readability.

Can I use this for general development?

Yes, while optimized for the TML-bench repository, its rigorous staging rules make it an excellent utility for any developer wanting to avoid 'dirty' commits.

Does the Checkpoint skill push code to remote repositories?

No, the skill is designed for local safety and explicitly blocks commands like git push and git rebase to ensure the user retains final control.

Git Safe Checkpoint

Name: Git Safe Checkpoint
Author: MykolaPinchuk

byMykolaPinchuk

データサイエンスとML

Automates secure, structured Git commits by enforcing guardrails against staging sensitive data, large datasets, and machine learning artifacts.

概要

The Checkpoint skill streamlines development within the TML-bench environment by providing a robust framework for saving progress safely. It eliminates the risk of accidental data leaks or repository bloat by automatically excluding competition data, run artifacts, SQLite databases, and sensitive credentials. The skill handles the entire commit lifecycle—from pre-flight status checks and selective staging of configuration files to the generation of structured commit messages that include unique agent identifiers—ensuring a clean, traceable, and secure version history.

主な機能

Pre-commit and post-commit status reporting for full transparency
0 GitHub stars
Automatic exclusion of competition data, runs, and large binary artifacts
Selective staging limited to source code, documentation, and config files
Security guardrails against committing secrets, PEM keys, and ENV files
Structured commit messages with Agent ID and functional area tagging

ユースケース

Creating safe progress markers during complex machine learning experiments
Automating agent-driven code changes with standardized commit logs
Maintaining repository hygiene in collaborative data science environments

概要

主な機能

Pre-commit and post-commit status reporting for full transparency
0 GitHub stars
Automatic exclusion of competition data, runs, and large binary artifacts
Selective staging limited to source code, documentation, and config files
Security guardrails against committing secrets, PEM keys, and ENV files
Structured commit messages with Agent ID and functional area tagging

ユースケース

Creating safe progress markers during complex machine learning experiments
Automating agent-driven code changes with standardized commit logs
Maintaining repository hygiene in collaborative data science environments