What problem does it solve? Executing a multi-task implementation plan in a single context often leads to lost focus, context bloat, and unverified results. This Skill orchestrates focused subagent workers with self-contained briefs so each task gets isolated execution context while the controller verifies actual outcomes against acceptance criteria. ## Core Features & Use Cases - Outcome-Oriented Delegation: Dispatch workers with self-contained briefs covering goals, acceptance criteria, owned files, and constraints, while allowing bounded evidence-based equivalent implementations. - Evidence-Based Verification: Inspect actual changed files and run proportionate checks instead of relying on status labels, mandatory review gates, or universal test transcripts. - Git and Safety Boundaries: Enforce strict rules so workers never commit, push, or perform Git writes without explicit user authorization, and escalate decisions affecting scope, security, or public APIs. - Use Case: You have an approved plan with five independent tasks. Dispatch focused workers per task or wave, integrate their returned diffs, run targeted verification, and request review only where risk warrants it. ## Quick Start Use the deepwork-subagent-driven-development skill to execute the approved implementation plan by dispatching focused workers for each independent task.
