moo-tasks

Record a human user request verbatim as an overarching Goal with rich Markdown PRD/spec. Tasks will link back to this goal. Caps open tasks to prevent over-planning.

Get full details of a goal including rich markdown PRD/spec, status metrics, and all child tasks.

Update goal details: rich markdown PRD/specification, title, verbatim prompt, max open tasks cap, or status.

List all project goals and their statuses (active, completed, dropped).

Check goal coverage, open tasks vs cap, loose ends (uncompleted tasks), and completion metrics.

Drop an entire goal and cascade drop all open tasks under it in one action with a mandatory reason.

Reopen a dropped or completed goal and optionally reopen its tasks.

Create a task under a goal (or standalone) with detailed technical description, acceptance criteria, type, tags, priority, declared files, and dependencies.

Batch create multiple tasks under a goal with full technical descriptions, criteria, types, tags, and dependencies in a single operation.

Update task properties: title, description, type, tags, priority, acceptance criteria, declared files, goal, or deferred state.

Link one or more prerequisite blocker tasks to a task with cycle validation.

Unlink a prerequisite blocker task from a task.

Auto-surface the next unblocked, highest-priority task ready for execution from the active ready queue, with optional file conflict avoidance for parallel agent swarms.

Retrieve full task details, subtasks, dependencies, notes, and evidence.

List and filter tasks by goal, status, priority, type, tag, agent, deferred state, or search text.

Claim a task exclusively before starting work. Verifies lease, agent concurrency limit, loop count, and file touch conflicts.

Renew active lease on claimed task while performing long-running work.

Voluntarily release a claimed task back to the todo queue with optional progress notes.

Handoff an in-flight claimed task from one agent to another with handoff summary.

Close a task with mandatory proof of work: commands run, output logs, modified files, and test proofs. Optionally auto-claim next unblocked task.

Atomically complete a task with proof and immediately claim the next unblocked ready task under the same goal in a single LLM tool call.

Mark a task verified done by human or verification subagent.

Reject a closed task with a mandatory reason, moving it back to todo and incrementing reopen counter.

Escalate a question/blocker to the human user with optional selectable choices and transition task to waiting-on-human.

Get the queue of all tasks currently waiting on human answers or decisions.

Provide an answer to a task waiting on human guidance, automatically resuming it into the ready queue.

Capture unexpected work found mid-task without dropping current claim. Can be marked must-fix-now (blocker) or deferred.

Append an immutable, timestamped note to a task (e.g. what was tried, reason for failure, architecture insight).

List all timestamped notes and attempt logs for a task.

Log a structured attempt failure note with error snippet, failure category, hypothesis, and next plan. Increments task attempt counter and automatically escalates to human if loop threshold is exceeded.

Drop a task with a mandatory reason (no approval step required).

Reopen a done or dropped task without losing history, incrementing reopen counter.

Undo the last status transition for a task, rolling back to its previous state.

Drop multiple tasks in a single operation with a shared mandatory reason.

Reopen multiple done or dropped tasks back into the todo queue.

Record a project-level architectural decision with rationale that outlives individual tasks.

List project architectural decisions so settled questions stay settled during planning.

Supersede an existing architectural decision with a new choice and rationale.

Merge two tasks that turn out to be duplicates, moving dependencies and subtasks.

Where-did-I-leave-off overview on session start: surfaces abandoned doing tasks, waiting-on-human, ready queue, settled decisions, and orphan tasks.

Ultra-dense token-optimized context block with active goal, claimed task, acceptance criteria, settled decisions, and file locks. Ideal for system prompt injection.

Pre-check if specific files are currently locked by another agent holding an active claim, preventing edit collisions.

Full-text ranked search (SQLite FTS5) across tasks, acceptance criteria, and architectural decisions (ADR).

Retrieve file-centric historical context: active file locks, past completed tasks, recent notes, and relevant architectural decisions for specific files before modifying code.

Fast-path vibe coding tool: Atomically creates a task under a goal and claims it exclusively in a single round-trip, setting declared files, type, tags, and lease duration.

Fast progress checkpoint: Appends a timestamped progress note to the in-flight task and optionally extends lease heartbeat in one call.

Parse a markdown plan, PRD, or task checklist into a Goal and atomic Tasks with dependencies.

Export all goals, tasks, notes, and decisions in Markdown, JSON, or Plain Text format.

Archive completed and dropped tasks out of the active working list.

List all registered project workspaces in the global Moo Tasks registry with their paths and git remotes.

Get details of the active workspace or a specific workspace by ID, name, or folder path.

Register a new project directory as a workspace in the global Moo Tasks database.

Update a workspace's display name or git remote URL.

Unregister/delete a workspace from the global registry.