Runs parallel blind dual-judge adversarial code reviews with fix and re-judgment cycles.
npx skills add https://github.com/jhannka/php-skills --skill judgment-day-jhannka
Please help me install this Agent Skill. Skill: judgment-day Source: https://github.com/jhannka/php-skills/tree/main/skills/judgment-day Command: npx skills add https://github.com/jhannka/php-skills --skill judgment-day-jhannka
High-intent search queries and answers about installing and using this skill.
Trigger the protocol by asking for a judgment day review on a specific target. Two blind judge sub-agents are launched in parallel with identical criteria, their findings are synthesized into a verdict table, and confirmed issues are fixed and re-judged.
Blind parallel review means two agents review the same code independently without knowing about each other. Findings confirmed by both are high-confidence, while single-judge findings are flagged as suspects, reducing individual reviewer blind spots.
A target is approved when both judges return clean, or after Round 2 when zero confirmed CRITICALs and zero confirmed real WARNINGs remain. Theoretical warnings and suggestions may remain without blocking approval.
After two fix iterations with remaining issues, the protocol asks the user whether to continue iterating. If the user declines, the review is marked ESCALATED with a report of remaining issues for manual human review.
Avoid it for trivial changes where two review rounds cost more than the risk of a bug, or when the target scope is unclear. The protocol requires a specific, well-defined target before launching judges.
wise-clawLoad project-specific Trellis development guidelines before starting implementation.
everquintEnforces trunk-based git workflow covering branches, worktrees, commits, PRs, reviews, releases, and rollbacks.
dobachiReview source code for quality, security, performance, and maintainability issues.
stablyaiReviews code changes, fixes issues in two rounds, then creates and merges a PR.
