メニュー
Gemini-native multi-model adversarial code review — iterative convergence loop for PRs and plans
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Gemini-native AI slop detection — scans all repos for AI-generated anti-patterns with 2M+ token grounding
Gemini-native autonomous close-out orchestrator — 20-step pipeline with whole-project grounding and hard gates
Gemini-native morning investigation pipeline — syncs repos, checks infra, and dispatches parallel probes with 2M+ token grounding
Gemini-native self-healing CI pipeline — autonomous fix loop with whole-project grounding
Gemini-native organization-wide CI recovery — scans all open PRs, classifies failures, and applies remediation with 2M+ token grounding
Gemini-native per-PR CI monitoring — polls GitHub Actions, auto-fixes introduced failures, and reports state with 2M+ token grounding
| name | hostile-reviewer |
| description | Gemini-native multi-model adversarial code review — iterative convergence loop for PRs and plans |
| capability | skill.hostile_reviewer |
| node_name | node_skill_hostile_reviewer_orchestrator |
| dot_color | 208 |
Multi-model adversarial review with iterative convergence, leveraging Gemini's 2M+ token window to perform "whole-project" reviews that identify structural flaws and architectural drift.
--pr <N>: PR number to review.--file <path>: Plan file to review.--models: Comma-separated model list (default: codex,deepseek-r1).--passes: Fixed number of passes (default: iterates to convergence).