What problem does it solve? Before running implementation on an OpenSpec change request (ZNИ), teams need an independent quality gate that confirms the proposal, design, tasks, and specs are coherent, traceable, and actually implementable — this Skill answers the question "can I safely run apply?" with a clear verdict. ## Core Features & Use Cases - Five-layer verification pipeline: Hygiene auto-fixes, Internal Coherence (quality controller + code-truth grep), Loop Detection for acceptance cycles, Problem-Solution Trace (Why → Requirements → Scenarios → Slices → Tasks), Independent Challenge (adversarial architect audit), and Implementation Readiness. - Binary GO/NO-GO verdict with chat-friendly output: Internal layer statuses are translated into plain verdict-card language for the user, with decision cards when a fork requires user choice and a silent re-run filter when nothing changed since the last report. - Depth modes and repair loop: Supports full, incremental, and lite verification depths with guardrails, plus an internal repair loop that fixes deterministic spec gaps without user interruption. - Use Case: A 1C developer finishes drafting a change with /opsx:new and runs /opsx:verify <name>; the skill checks slice coherence, runs an adversarial design challenge, auto-fixes checkbox hygiene, and returns "можно запускать apply" with a dated verification report. ## Quick Start Ask the assistant to verify an OpenSpec change by name, for example: run /opsx:verify on my change to check whether it is safe to start implementation.
