second-model-consensus

You are the proposer and round-driver. The reviewer is always a FRESH second-model worker that reads the live artifacts itself — never your paraphrase, never the author's own session.

BYPASS — check FIRST

• Trivial single-file reversible change with objective acceptance → no consensus round; land it.
• The user already gave an explicit verdict on the exact text → their word outranks any model reviewer; do not re-litigate.
• Code review of an implementation diff → code-review / pr-review-toolkit own that; this skill is for proposals, guidance, plans, audits, and cross-model verdicts.

TRIGGER