work

Parallel task execution is coordinated by spawning worker agents that resolve dependencies and claim ready tasks from the backlog. Worker loops manage task orchestration, handle recovery, detect cycles, and avoid conflicts during concurrent processing.

Dependency-aware task orchestration uses a verification gate where each worker result is checked by a verifier before committing changes. If verification fails, the task is automatically re-dispatched up to two times for recovery.

New tasks promoted to the todo list during parallel execution runs are automatically discovered and picked up by worker agents. This dynamic discovery mechanism ensures newly added backlog items are processed without interrupting the active orchestration loop.

Managing task conflicts and cycle detection in automated workflows is handled by the parallel worker loop during task execution. The orchestration mechanism automatically detects dependency cycles and avoids conflicts when multiple worker agents claim and process ready tasks.

Parallel worker orchestration does not support modeling or design tasks, as its scope is strictly limited to executing refined backlog tasks. Modeling responsibilities are left to other specialized skills, while this tool focuses purely on task execution and verification.

Dependency-aware worker loops should be used when tasks have explicit dependencies, require coordinated parallel execution, and must pass verification before committing changes. This approach replaces manual scheduling by automatically resolving task dependencies and managing recovery.