What problem does it solve?
Long-running Codex work across multiple sessions can lose context, require handoffs, and risk progress decay. This Skill provides a structured autonomous-task framework using an Initializer + Executor pattern with persistent state under .autonomous to resume, continue, and manage multi-session tasks without losing progress.
Core Features & Use Cases
- Dual-agent orchestration: Split work between an INITIALIZER (breakdown) and an EXECUTOR (progressive execution) with automatic session continuation.
- Session persistence: Stores task state under .autonomous/<task-name>/ to support resumable tasks across Codex sessions.
- On-demand task data: Patterns for running tasks deterministically, resuming, listing tasks, and separating task tracking from project files.
- Use Case: Long-running experiments, multi-session data gathering, or complex analyses that need pause/resume without losing state.
Quick Start
Start a new autonomous task by running the session runner with a description of the task.