What problem does it solve? Running multi-turn coding tasks with an LLM agent often loses context, lacks safety controls, and makes it hard to recover from bad changes. This Skill operates oh-my-cli, a CLI code agent that reads and writes files, executes shell commands, and manages sessions with approval gates, undo/redo checkpoints, and context compaction. ## Core Features & Use Cases - Autonomous Coding with Safety Gates: The agent uses file and shell tools with approval gates for mutations, plus atomic per-turn checkpoints so changes can be previewed, undone, or redone. - Session Persistence and Compaction: Sessions are stored as JSONL transcripts that can be resumed, browsed, exported to Markdown, and compacted when they approach provider context limits. - Multi-Profile Model Configuration: Switch between hosted and local OpenAI-compatible models via profiles in a user settings file that references API keys through environment variables instead of storing raw credentials. - Use Case: Ask the agent to refactor a module, preview the undo plan with a dry run if the result is wrong, then export the session transcript for a teammate to review exactly what the agent changed. ## Quick Start Use oh-my-cli to start an autonomous code agent session that refactors the auth module in this project, and show me how to undo the turn if the changes are wrong.
