What problem does it solve? Letting an AI coding agent work directly on your real repository risks unwanted or destructive changes. This Skill isolates the agent on a disposable clone, optionally inside a Podman or Docker container, so you review the diff and decide what to apply. ## Core Features & Use Cases - Sandboxed Execution: Clones the target repo into a task directory and delegates the coding work to a subagent that cannot touch the original repo. - Optional Container Isolation: Auto-detects Podman or Docker to run builds and tests inside a pre-built sandbox image with Node.js, Python, and git. - Review-Gated Apply: Captures git diffs per task and applies them only after explicit confirmation, with support for applying all changes or filtering by file glob. - Use Case: Ask the agent to fix a login bug in your project; it clones the repo, makes and tests changes in the sandbox, then shows you a diff you can inspect, apply, or discard. ## Quick Start Run yolo with a task name and repo path, for example: yolo run fix-login-bug ~/projects/myapp "Fix the OAuth redirect loop in login".