myna-dev-execute-tasks

To automate a dev task queue, you define pending work in a tmp/tasks.md file, and the skill orchestrates execution in FIFO order on a single fix branch, delegating each task to a subagent and creating a single pull request.

A single-branch pull request workflow handles multiple sequential tasks by recording the base commit for each task, delegating work to per-task subagents, and bundling all commits into one PR to reduce context switching and handoffs.

Yes, you can add new tasks during an active task orchestration run. The queue re-scans for mid-run task discovery before processing each item, ensuring newly added tasks are included in the current execution cycle.

The best way to maintain run logs and update a changelog during task automation is to use an orchestration skill that automatically writes per-run logs to a tmp/[date]/run.md file and applies configured changelog updates as tasks complete.

Yes, per-task review and commit protocols work for both feature work and bug fixes. The deterministic task queue delegates each item to a subagent, reviews the specific commit, and reports results before proceeding.

Orchestrating a deterministic dev task queue requires a tmp/tasks.md input file containing the pending tasks. The skill then reads this file to process all tasks sequentially on a fix/[date] branch.