What problem does it solve? Engineering teams delegating implementation work to AI agents often lack a disciplined process for defining completion criteria, verifying output quality, and controlling model costs, leading to unreviewed code and wasted tokens. ## Core Features & Use Cases - Eval-First Execution: Define capability and regression evals before implementation, capture baseline failure signatures, and compare deltas after changes. - 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. - Cost-Aware Model Routing: Route tasks by complexity across Haiku (classification, narrow edits), Sonnet (implementation, refactors), and Opus (architecture, root-cause analysis), escalating tiers only on clear reasoning gaps. - Use Case: When refactoring a multi-module service with AI agents, decompose the work into verifiable units, run baseline evals, assign Sonnet to implementation units, and review only invariants, error boundaries, and auth assumptions. ## Quick Start Ask the agent to plan and execute your next engineering task using the agentic-engineering workflow with eval-first verification and tiered model routing.
