What problem does it solve? When AI agents perform most implementation work, teams lack a disciplined process for defining completion criteria, verifying output quality, and controlling model costs. This Skill provides an operating framework for agentic engineering so humans can enforce quality and risk controls over AI-generated code. ## Core Features & Use Cases - Eval-First Loop: Define capability and regression evals, capture baseline failure signatures, then re-run evals after implementation to compare deltas. - Task Decomposition: Apply the 15-minute unit rule so each work unit is independently verifiable with a single dominant risk and a clear done condition. - Model Routing & Cost Discipline: Route tasks across Haiku, Sonnet, and Opus tiers by complexity, and track model, tokens, retries, and success rates per task. - Use Case: A team delegating a multi-file refactor to an AI agent uses this Skill to split the work into verifiable units, assign Sonnet for implementation and Opus for architecture review, and validate results with regression evals before merging. ## Quick Start Use the agentic-engineering skill to plan and execute my next AI-assisted implementation task with eval-first verification and model routing.
