Skill PR for review: judgment-hygiene-stack (#733)

[SyntagmaNull]

Hi — I submitted PR #733 about two months ago and it hasn't had any review yet, so flagging it here.
It's a tri-skill pipeline (judgment-hygiene-stack) that addresses structural failure modes in Claude's output: premise-smuggling, verification bypass, synthetic completion, and layer collapse. The pipeline enforces an execution order — structural routing first, conditional verification second, judgment construction last — so downstream stages can't override upstream routing decisions.
The skill was developed through extended real-usage iteration across Claude, GPT, and Gemini, and tested across mixed fact/interpretation prompts, multimodal conflict, and high-emotion inputs.
PR: #733
Happy to answer questions or adjust anything based on feedback.