Optimizes agent-driven development through eval-first execution, task decomposition, and cost-aware model routing.
This skill transforms Claude into a disciplined agentic engineer by implementing a structured workflow centered on pre-defined completion criteria and regression testing. It optimizes resource utilization through intelligent model tier routing—leveraging Haiku for narrow edits, Sonnet for implementation, and Opus for architecture—while enforcing a 15-minute unit rule for task decomposition. By focusing human oversight on invariants and security rather than style, it streamlines the software development lifecycle for maximum autonomy, reliability, and cost-efficiency.
Key Features
010 GitHub stars
0215-minute unit decomposition for verifiable work blocks
03Eval-first implementation loop with regression signature capturing
04Strategic session management and milestone-based compacting
05Risk-focused review process targeting invariants and security
06Cost-aware model routing across Haiku, Sonnet, and Opus
Use Cases
01Scaling complex feature development with autonomous AI agents
02Optimizing API costs by routing tasks to the appropriate model intelligence tier
03Ensuring code quality in large-scale refactors via regression checks
