universal-tasks

Multi-session task coordination across agents is managed through a persistent, dependency-aware task system using file-based storage. It tracks task IDs, statuses, and dependencies to synchronize work across multiple models and sessions without losing progress.

Handling task dependencies in long-running Python scripts relies on a persistent task schema that validates blockers and assigned models. Included Python scripts enforce atomic claiming and update task statuses to prevent duplicate effort during execution.

Abandoned tasks during cross-session agent orchestration are recovered through built-in stale-claim recovery. This automatic mechanism resets inactive task claims, ensuring that blocked or long-running architectural changes resume quickly without manual intervention.

File-based storage supports multi-model agent collaboration by maintaining a persistent JSON task list. It enables different models, such as Opus and Flash, to atomically discover, claim, and update tasks across separate contexts and sessions.

Preventing duplicate work when multiple agents claim tasks simultaneously requires atomic discovery and claiming logic. Provided Python scripts validate task IDs and lock task statuses atomically, ensuring only one agent modifies a specific task at a time.

Tracking cross-session project dependencies requires no specific framework, only file-based storage and Python scripts. You initialize a JSON task list and use the provided scripts to validate, claim, and complete tasks without external database dependencies.