What problem does it solve? Engineering teams and AI agents lack a shared, portable set of judgment standards for deciding how much rigor a change deserves, what counts as delivered work, and when to escalate decisions to a human. This Skill encodes those preferences so they apply in any project, independent of a specific agent loop or tooling configuration. ## Core Features & Use Cases - Two-tier rigor model: Distinguishes a non-negotiable quality floor (regression coverage, observability, security-by-design) from risk-calibrated judgment that scales with blast radius. - Twelve working principles: Covers plan-first design, thin vertical slices, surgical changes with tracked debt, and dive-deep investigation discipline. - Delivery vs hygiene accounting: Defines delivered work as user-visible change and requires reporting product slices against hygiene slices each session. - Escalation standard: Specifies a five-clause test and a terse, options-based form for raising decision pendencies versus action pendencies to a human. - Use Case: An agent working in an unfamiliar repository uses these principles to decide whether a refactor needs human sign-off, whether a green CI check is trustworthy evidence, and how to escalate a scope trade-off. ## Quick Start Ask the agent to apply the engineering standards to decide how much rigor a proposed change deserves and whether it should be escalated.
