task-agent 0.1.90

A prioritized, file-based task queue for autonomous agentic workers

Task Agent 🤖

A prioritized, file-based task queue for autonomous agentic workers. This system uses a "Mission Control" approach to manage multiple agents working on git-tracked improvements across various branches and worktrees.

📂 System Architecture

The project follows a specific folder structure to manage the lifecycle of an improvement:

• docs/issues/: The core queue.
  • mission.usv: A prioritized list of issues using Unit Separator Value (\x1f) format. This serves as the source of truth for task priority.
  • datapackage.json: Frictionless Data metadata for the mission.usv schema.
  • pending/: New issues awaiting triage or assignment.
  • draft/: Issues currently being refined or planned.
  • active/: Issues actively being worked on by an agent.
  • completed/{year}/: Successfully implemented and verified improvements.
• .gwt/: Git Worktree directory where active branches are checked out for isolated agent execution.

🛠️ Tooling: ta (Task Agent CLI)

The ta tool automates the transition of issues through the queue and manages the underlying git infrastructure.

Commands

Command	Action
ta next	Displays the top prioritized issue from mission.usv.
ta new	Creates a new issue file and adds it to the queue.
ta done	Moves issue to completed/, removes it from the queue, and auto-commits the result.
ta start <slug>	Moves issue to active/, creates a git branch, and sets up a .gwt/ worktree.
ta run <slug>	Invokes the sidecar worker defined at .ta/worker to process an active issue.

🚀 Workflow

1. Prioritize: Use ta new -t "Title" to add a task.
2. Review: Run ta next to see what is currently at the top of the queue.
3. Dispatch: (Planned) Run ta start <slug> to prepare the workspace.
4. Execute: (Planned) The agent (Gemini CLI) processes the task in its isolated worktree via ta run <slug>.
5. Finalize: Once verified, ta done moves the task to the finished state.

🤖 Gemini CLI Integration

By utilizing the Gemini CLI's headless mode, this system can orchestrate multiple agents simultaneously without requiring manual terminal management, significantly improving portability and scalability between Windows and WSL environments.