What problem does it solve? Single-reviewer code or document reviews can miss defects or produce unverified claims. This Skill raises confidence on high-stakes targets by having two blind judges review the same frozen target independently, treating their agreement as corroboration and escalating contradictions to you. ## Core Features & Use Cases - Dual blind review: Launches two independent judge subagents with byte-identical prompts so neither sees the other's work before merging. - Convergence-based merge: Classifies findings as confirmed (both judges), suspect (one judge), contradiction (incompatible claims), or info, with a structured verdict of APPROVED or ESCALATED. - Code and artifact modes: Reviews frozen diffs, branches, PRs, or commit ranges, as well as single documents like specs, designs, plans, and RFCs. - Optional ledger and re-judgment: Persists a review ledger at docs/reviews/{target-slug}.md on request and supports up to two scoped re-judgments after user-applied fixes. - Use Case: Before merging a critical pull request, invoke the protocol to get two independent reviews; only defects both judges confirm enter the fix loop, while solitary findings stay flagged as unproven suspects. ## Quick Start Ask the assistant to run judgment-day on a specific pull request, commit range, or document you want independently double-reviewed.
