What problem does it solve? Teams lose control when AI agents claim tasks, edit code, or mark work Done without verified approval, atomic ownership, or real evidence. This Skill enforces a disciplined task lifecycle: gate checks before any write, safe claim mechanisms, scoped execution, two-axis review, and evidence-backed completion. ## Core Features & Use Cases - Gate-controlled execution: Verifies branch-specific approvals (Spike, Bounded, Feature G1–G4) before any source, test, config, or migration changes. - Three-tier subagent workflow: Dispatches Task Worker, Task Reviewer, and an integration-level Reviewer for parallel task execution with canonical task records. - Evidence-based completion: Requires revision-specific test evidence, two-axis review (Spec Compliance and Standards), and Browser Native checkpoints for UI changes before Done. - Use Case: A team lead asks the agent to implement an approved feature; the Skill verifies G1–G4 approvals, claims the task atomically, delegates work to subagents, collects review findings and test evidence, and only then updates the backlog and marks the task Done. ## Quick Start Ask the agent to claim and execute an approved task from the roadmap, then review and verify it with evidence before marking it Done.
