What problem does it solve? AI coding agents often repeat the same mistakes, lose track of goals during long sessions, and mark tasks complete without verification. This Skill provides a structured set of engineering principles that make agent behavior more consistent, verifiable, and self-improving. ## Core Features & Use Cases - Plan-First Workflow: Enforces plan mode for non-trivial tasks with checkable items tracked in tasks/todo.md. - Self-Improvement Loop: Converts every user correction into a permanent rule in tasks/lessons.md so mistakes are not repeated. - Subagent Delegation & Verification Gates: Guides when to offload work to subagents and requires test evidence before marking tasks done. - Use Case: When writing an AGENTS.md or CLAUDE.md for a new repository, apply these principles to define the agent's planning, verification, and lesson-capture standards from day one. ## Quick Start Ask the agent to draft an AGENTS.md file for your project using these engineering principles, including a verification checklist and a lessons-file convention.
