What problem does it solve?
Coordinating multiple AI workers on a shared project often leads to duplicated work, unclear ownership, hidden dependencies, and noisy status updates. This Skill gives a Leader agent a structured method to decompose work into bounded tasks, assign single owners, choose the right execution mode, and decide when results are truly accepted.
Core Features & Use Cases
- Execution Mode Selection: Choose between DAG mode for finite, plannable dependency graphs and Loop mode for iterative work that repeats until a stop condition or quality gate is met.
- Task Boundary Design: Define tasks with one owner, one inspectable output, clear acceptance criteria, and deliverables isolated in shared/tasks/{task-id}/ to prevent write conflicts.
- Result Acceptance Decisions: After checking a task, decide whether to accept it, request revision, add a verifier task, ask the requester, or report a blocker, so only accepted results unblock downstream work.
- Use Case: A Leader receives a request to iterate on a research report until it passes review. The Skill guides the Leader to pick Loop mode, plan one iteration at a time, evaluate each round, and report only accepted outcomes to the requester.
Quick Start
Act as the Team Leader and break this project into bounded tasks with clear owners, choosing DAG or Loop mode and defining acceptance criteria before delegating.
