{"id":51223010,"url":"https://github.com/baiyanmo/claude-code-multi-model-review","last_synced_at":"2026-07-06T19:00:42.726Z","repository":{"id":365413774,"uuid":"1271985377","full_name":"baiyanmo/claude-code-multi-model-review","owner":"baiyanmo","description":"Multi-model code review skill for Claude Code — invoke DeepSeek, Doubao, Qwen, OpenAI \u0026 more in parallel to review your git diff. 多模型并行代码审查 skill，同时调用多个 AI 模型交叉验证。","archived":false,"fork":false,"pushed_at":"2026-06-17T08:06:28.000Z","size":25,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"master","last_synced_at":"2026-06-17T09:20:46.311Z","etag":null,"topics":["ai-review","claude-code","claude-code-skill","code-review","cross-validation","deepseek","devtools","doubao","git-diff","multi-model","openai","qwen"],"latest_commit_sha":null,"homepage":null,"language":"JavaScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/baiyanmo.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-06-17T07:23:57.000Z","updated_at":"2026-06-17T08:30:39.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/baiyanmo/claude-code-multi-model-review","commit_stats":null,"previous_names":["baiyanmo/claude-code-multi-model-review"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/baiyanmo/claude-code-multi-model-review","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/baiyanmo%2Fclaude-code-multi-model-review","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/baiyanmo%2Fclaude-code-multi-model-review/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/baiyanmo%2Fclaude-code-multi-model-review/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/baiyanmo%2Fclaude-code-multi-model-review/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/baiyanmo","download_url":"https://codeload.github.com/baiyanmo/claude-code-multi-model-review/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/baiyanmo%2Fclaude-code-multi-model-review/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":35202786,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-07-06T02:00:07.184Z","response_time":106,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai-review","claude-code","claude-code-skill","code-review","cross-validation","deepseek","devtools","doubao","git-diff","multi-model","openai","qwen"],"created_at":"2026-06-28T09:00:31.116Z","updated_at":"2026-07-06T19:00:42.672Z","avatar_url":"https://github.com/baiyanmo.png","language":"JavaScript","funding_links":[],"categories":["Source Catalog","CLI \u0026 Local Workflows"],"sub_categories":[],"readme":"# Claude Code Multi-Model Review Skill\n\n**[English](./README.md#en) | [中文](./README.md#zh) | [日本語](./README.md#ja) | [한국어](./README.md#ko)**\n\n![GitHub Stars](https://img.shields.io/github/stars/baiyanmo/claude-code-multi-model-review?style=flat\u0026color=yellow)\n![GitHub License](https://img.shields.io/github/license/baiyanmo/claude-code-multi-model-review?color=blue)\n![Node Version](https://img.shields.io/badge/node-%3E%3D14-brightgreen)\n![Platform](https://img.shields.io/badge/platform-win%20%7C%20mac%20%7C%20linux-lightgrey)\n\n---\n\n\u003ca id=\"en\"\u003e\u003c/a\u003e\n## English\n\nCall multiple external AI models to review your code changes in parallel. Each model gives independent feedback, then the results are compared side by side.\n\n### Why\n\nClaude Code's built-in model may route through a proxy to a single backend model (e.g., DeepSeek), limiting review diversity. This skill lets you invoke models from different providers (DeepSeek, Doubao, Qwen, OpenAI, etc.) simultaneously on the same diff — catching blind spots that any single model might miss.\n\n### Features\n\n- **Parallel reviews** — All selected models run at the same time, no waiting\n- **Independent opinions** — Each model's output is preserved in full, no truncation\n- **Cross-validation** — Issues flagged by multiple models have higher confidence\n- **Divergence comparison** — See where models disagree\n- **Easy to extend** — Add new model APIs anytime\n\n### Demo\n\nRunning a review on a README change (adding shield badges):\n\n```\n[review.js] 正在调用 豆包 DeepSeek (ep-20260607142413-sd26g)...\n[review.js] 豆包 DeepSeek 返回完成\n\n## 豆包 DeepSeek (Volcengine)\n\n### Overall Assessment\nThis is a clean, trivial README improvement change with no critical/high/medium\nissues, no functional bugs, security risks or performance problems at all.\n\n### What this change does well\n1. The 4 added standard shields.io badges are placed logically right after the\n   language navigation bar, making core project metadata immediately visible.\n2. All badge color choices follow common open source conventions.\n\n### Minor Low-Severity Findings\n| Severity | File       | Description                                   |\n|----------|------------|-----------------------------------------------|\n| LOW      | README.md  | Node Version badge is static, won't auto-sync |\n\nWith 2+ models, you get a cross-validation table showing which issues were\nflagged by multiple models — higher confidence findings you can act on first.\n```\n\n### File Structure\n\n```\ncustom-review/\n  SKILL.md              # Workflow orchestration (Claude Code reads this)\n  review.js             # Single-model review script (OpenAI-compatible API)\n  models.example.json   # Example model config template\n  models.json           # Your actual model config (gitignored)\n```\n\n### Quick Start\n\n#### 1. Install\n\nCopy the `custom-review/` directory to `~/.claude/skills/`:\n\n```bash\ngit clone https://github.com/baiyanmo/claude-code-multi-model-review.git\ncp -r claude-code-multi-model-review ~/.claude/skills/custom-review\n```\n\n#### 2. Configure Models\n\nIn Claude Code, say:\n\n\u003e 添加审查模型\n\nOr in English:\n\n\u003e Add a review model\n\nProvide the following info for each model:\n- Name (e.g., \"DeepSeek\")\n- API Base URL (OpenAI-compatible format)\n- API Key\n- Model ID (e.g., `deepseek-v4-pro`)\n\nConfigure at least 2 models to get the most out of multi-model review.\n\n#### 3. Review Code\n\nIn a git repo, say:\n\n\u003e 审查代码 / review / check my changes\n\nThe skill triggers, lets you pick which models to use, then runs them all in parallel.\n\n### Supported Providers\n\nAny OpenAI-compatible API works, including:\n\n| Provider | Base URL |\n|----------|----------|\n| DeepSeek | `https://api.deepseek.com/v1` |\n| Alibaba Bailian (Qwen) | `https://dashscope.aliyuncs.com/compatible-mode/v1` |\n| Volcengine (Doubao) | `https://ark.cn-beijing.volces.com/api/v3` |\n| OpenAI | `https://api.openai.com/v1` |\n| SiliconFlow | `https://api.siliconflow.cn/v1` |\n| Custom proxy | Any URL |\n\n### CLI Usage\n\nYou can also run `review.js` directly from the terminal:\n\n```bash\nnode review.js --diff /tmp/my_diff.txt --model deepseek --config models.json\nnode review.js --diff /tmp/my_diff.txt --model doubao --angle \"Focus on security issues\" --config models.json\n```\n\n### Security\n\n- `models.json` contains your API keys — excluded via `.gitignore`\n- `review.js` uses only Node.js built-in modules (no `npm install` needed)\n- API calls go over HTTPS, timeout at 120 seconds\n\n---\n\n\u003ca id=\"zh\"\u003e\u003c/a\u003e\n## 中文\n\n调用多个外部 AI 模型并行审查代码改动，每个模型给出独立意见，综合对比各模型的发现。\n\n### 为什么需要\n\nClaude Code 本身的模型可能通过代理指向同一个底层模型（如 DeepSeek），审查视角单一。这个 skill 让你同时调用不同厂商的模型（DeepSeek、豆包、千问、OpenAI 等）审查同一份 diff，互相验证，避免单一模型的盲区。\n\n### 功能\n\n- **多模型并行审查** — 同时调用多个外部模型，互不等待\n- **独立意见** — 每个模型用自己的角度评审，完整保留不做删减\n- **交叉验证** — 多个模型同时发现的问题置信度最高\n- **分歧对比** — 展示不同模型之间的意见分歧\n- **自由扩展** — 随时添加新的模型 API\n\n### 效果演示\n\n对 README 添加徽章的改动进行审查：\n\n```\n[review.js] 正在调用 豆包 DeepSeek (ep-20260607142413-sd26g)...\n[review.js] 豆包 DeepSeek 返回完成\n\n## 豆包 DeepSeek (Volcengine)\n\n### Overall Assessment\n这是一个干净的 README 改进，没有严重/高/中级别的问题，\n无功能缺陷、安全风险或性能问题。\n\n### 做得好的地方\n1. 4 个 shields.io 徽章放在语言切换栏下方，位置合理，让项目核心元信息一目了然\n2. 徽章配色符合开源惯例，动态徽章 URL 配置正确\n\n### 低严重度发现\n| 级别 | 文件       | 描述                                   |\n|------|------------|----------------------------------------|\n| 低   | README.md  | Node Version 徽章是静态的，升级后需手动更新 |\n\n配置 2 个以上模型时，还会生成交叉验证表 — 多个模型同时发现的问题置信度最高，\n值得优先处理。\n```\n\n### 文件结构\n\n```\ncustom-review/\n  SKILL.md              # Skill 编排流程（Claude Code 读这个）\n  review.js             # 单模型审查脚本（OpenAI 兼容 API）\n  models.example.json   # 模型配置示例模板\n  models.json           # 你的实际模型配置（gitignored）\n```\n\n### 快速开始\n\n#### 1. 安装 Skill\n\n将 `custom-review/` 目录复制到 `~/.claude/skills/` 下：\n\n```bash\ngit clone https://github.com/baiyanmo/claude-code-multi-model-review.git\ncp -r claude-code-multi-model-review ~/.claude/skills/custom-review\n```\n\n#### 2. 配置模型\n\n在 Claude Code 中说：\n\n\u003e 添加审查模型\n\n按提示填写以下信息：\n- 名称（自定义，如\"豆包\"）\n- API Base URL（OpenAI 兼容格式）\n- API Key\n- 模型 ID（如 `deepseek-v4-pro`、`ep-xxx`）\n\n建议至少配置 2 个模型才能发挥\"多模型\"的价值。\n\n#### 3. 审查代码\n\n在有 git 仓库的目录下说：\n\n\u003e 审查代码 / review / 帮我看看代码\n\nSkill 会触发，让你选择用哪几个模型，然后并行审查当前 diff。\n\n### 支持的模型\n\n任何 OpenAI 兼容 API 都可以，包括但不限于：\n\n| 厂商 | Base URL |\n|------|----------|\n| DeepSeek | `https://api.deepseek.com/v1` |\n| 阿里云百炼（千问） | `https://dashscope.aliyuncs.com/compatible-mode/v1` |\n| 火山引擎（豆包） | `https://ark.cn-beijing.volces.com/api/v3` |\n| OpenAI | `https://api.openai.com/v1` |\n| SiliconFlow | `https://api.siliconflow.cn/v1` |\n| 其他兼容代理 | 自定义 |\n\n### 命令行用法\n\n也可以在终端直接调用 review.js：\n\n```bash\nnode review.js --diff /tmp/my_diff.txt --model deepseek --config models.json\nnode review.js --diff /tmp/my_diff.txt --model doubao --angle \"重点关注安全问题\" --config models.json\n```\n\n### 安全\n\n- `models.json` 包含 API Key，已在 `.gitignore` 中排除\n- review.js 只依赖 Node.js 内置模块，无需 `npm install`\n- API 调用通过 HTTPS，超时 120 秒\n\n---\n\n\u003ca id=\"ja\"\u003e\u003c/a\u003e\n## 日本語\n\n複数の外部 AI モデルを並行実行し、コード変更をレビューします。各モデルが独立した意見を出し、結果を横断比較します。\n\n### なぜ必要か\n\nClaude Code のモデルがプロキシ経由で単一のバックエンドモデル（例: DeepSeek）にルーティングされている場合、レビューの視点が偏る可能性があります。この Skill を使うと、異なるプロバイダー（DeepSeek、豆包、Qwen、OpenAI など）のモデルを同じ diff に対して同時に実行し、単一モデルでは見逃す可能性のある問題を検出できます。\n\n### 機能\n\n- **並行レビュー** — 選択した全モデルが同時に実行、待ち時間なし\n- **独立した意見** — 各モデルの出力を完全に保持、削除なし\n- **クロスバリデーション** — 複数モデルが指摘した問題は信頼度が高い\n- **意見の分岐を比較** — モデル間の意見の相違を表示\n- **簡単に拡張** — いつでも新しいモデル API を追加可能\n\n### デモ\n\nREADME の変更（バッジ追加）をレビューした例：\n\n```\n[review.js] 正在调用 豆包 DeepSeek (ep-20260607142413-sd26g)...\n[review.js] 豆包 DeepSeek 返回完成\n\n## 豆包 DeepSeek (Volcengine)\n\n### Overall Assessment\nこれはクリーンな README の改善で、致命的・高・中の問題はなく、\n機能バグやセキュリティリスクもありません。\n\n### 良い点\n1. 4つの shields.io バッジが言語切り替えバーの下に適切に配置され、\n   プロジェクトのコア情報が一目でわかります\n2. バッジの配色は一般的なオープンソースの慣例に従っています\n```\n\n### ファイル構成\n\n```\ncustom-review/\n  SKILL.md              # ワークフロー定義 (Claude Code が読み込む)\n  review.js             # 単一モデルレビュースクリプト (OpenAI 互換 API)\n  models.example.json   # モデル設定のサンプルテンプレート\n  models.json           # 実際のモデル設定 (gitignored)\n```\n\n### クイックスタート\n\n#### 1. インストール\n\n`custom-review/` ディレクトリを `~/.claude/skills/` にコピーします：\n\n```bash\ngit clone https://github.com/baiyanmo/claude-code-multi-model-review.git\ncp -r claude-code-multi-model-review ~/.claude/skills/custom-review\n```\n\n#### 2. モデルの設定\n\nClaude Code で次のように言います：\n\n\u003e 添加审查模型\n\n各モデルの以下の情報を入力してください：\n- 名前（例: \"DeepSeek\"）\n- API Base URL（OpenAI 互換形式）\n- API Key\n- モデル ID（例: `deepseek-v4-pro`）\n\nマルチモデルレビューを最大限活用するために、最低 2 つのモデルを設定することをお勧めします。\n\n#### 3. コードレビュー\n\ngit リポジトリ内で次のように言います：\n\n\u003e 审查代码 / review / 帮我看看代码\n\nSkill が起動し、使用するモデルを選択すると、すべてのモデルが並行実行されます。\n\n### 対応プロバイダー\n\nOpenAI 互換 API であればすべて対応しています：\n\n| プロバイダー | Base URL |\n|------|----------|\n| DeepSeek | `https://api.deepseek.com/v1` |\n| Alibaba Bailian (Qwen) | `https://dashscope.aliyuncs.com/compatible-mode/v1` |\n| Volcengine (Doubao) | `https://ark.cn-beijing.volces.com/api/v3` |\n| OpenAI | `https://api.openai.com/v1` |\n| SiliconFlow | `https://api.siliconflow.cn/v1` |\n| カスタムプロキシ | 任意の URL |\n\n### CLI での使用\n\nターミナルから直接 `review.js` を実行することもできます：\n\n```bash\nnode review.js --diff /tmp/my_diff.txt --model deepseek --config models.json\nnode review.js --diff /tmp/my_diff.txt --model doubao --angle \"セキュリティ問題に注目\" --config models.json\n```\n\n### セキュリティ\n\n- `models.json` には API キーが含まれます — `.gitignore` で除外済み\n- `review.js` は Node.js 組み込みモジュールのみを使用（`npm install` 不要）\n- API 呼び出しは HTTPS 経由、タイムアウトは 120 秒\n\n---\n\n\u003ca id=\"ko\"\u003e\u003c/a\u003e\n## 한국어\n\n여러 외부 AI 모델을 병렬로 호출하여 코드 변경 사항을 검토합니다. 각 모델이 독립적인 의견을 제시하고, 결과를 나란히 비교합니다.\n\n### 필요한 이유\n\nClaude Code의 모델이 프록시를 통해 단일 백엔드 모델(예: DeepSeek)로 라우팅되는 경우, 리뷰 관점이 편향될 수 있습니다. 이 Skill을 사용하면 서로 다른 제공자(DeepSeek, 豆包, Qwen, OpenAI 등)의 모델을 동일한 diff에 대해 동시에 실행하여, 단일 모델이 놓칠 수 있는 문제를 발견할 수 있습니다.\n\n### 기능\n\n- **병렬 리뷰** — 선택한 모든 모델이 동시에 실행되어 대기 시간 없음\n- **독립적인 의견** — 각 모델의 출력을 완전히 보존, 삭제 없음\n- **교차 검증** — 여러 모델이 지적한 문제는 신뢰도가 높음\n- **의견 차이 비교** — 모델 간 의견 불일치를 표시\n- **쉬운 확장** — 언제든지 새 모델 API 추가 가능\n\n### 데모\n\nREADME 변경사항(배지 추가)을 리뷰한 예시:\n\n```\n[review.js] 正在调用 豆包 DeepSeek (ep-20260607142413-sd26g)...\n[review.js] 豆包 DeepSeek 返回完成\n\n## 豆包 DeepSeek (Volcengine)\n\n### Overall Assessment\n이것은 깨끗한 README 개선으로, 치명적/높음/중간 수준의 문제가 없으며\n기능 버그, 보안 위험, 성능 문제가 전혀 없습니다.\n\n### 잘한 점\n1. 4개의 shields.io 배지가 언어 전환 바 아래에 적절히 배치되어\n   프로젝트의 핵심 메타 정보를 한눈에 파악할 수 있습니다\n2. 배지 색상은 일반적인 오픈소스 관례를 따릅니다\n```\n\n### 파일 구조\n\n```\ncustom-review/\n  SKILL.md              # 워크플로우 정의 (Claude Code가 읽음)\n  review.js             # 단일 모델 리뷰 스크립트 (OpenAI 호환 API)\n  models.example.json   # 모델 설정 예제 템플릿\n  models.json           # 실제 모델 설정 (gitignored)\n```\n\n### 빠른 시작\n\n#### 1. 설치\n\n`custom-review/` 디렉토리를 `~/.claude/skills/`로 복사합니다:\n\n```bash\ngit clone https://github.com/baiyanmo/claude-code-multi-model-review.git\ncp -r claude-code-multi-model-review ~/.claude/skills/custom-review\n```\n\n#### 2. 모델 설정\n\nClaude Code에서 다음과 같이 말합니다:\n\n\u003e 添加审查模型\n\n각 모델에 대해 다음 정보를 입력하세요:\n- 이름 (예: \"DeepSeek\")\n- API Base URL (OpenAI 호환 형식)\n- API Key\n- 모델 ID (예: `deepseek-v4-pro`)\n\n멀티 모델 리뷰를 최대한 활용하려면 최소 2개 이상의 모델을 설정하는 것이 좋습니다.\n\n#### 3. 코드 리뷰\n\ngit 저장소에서 다음과 같이 말합니다:\n\n\u003e 审查代码 / review / 帮我看看代码\n\nSkill이 실행되어 사용할 모델을 선택하면, 모든 모델이 병렬로 실행됩니다.\n\n### 지원되는 제공자\n\nOpenAI 호환 API라면 모두 사용 가능합니다:\n\n| 제공자 | Base URL |\n|------|----------|\n| DeepSeek | `https://api.deepseek.com/v1` |\n| Alibaba Bailian (Qwen) | `https://dashscope.aliyuncs.com/compatible-mode/v1` |\n| Volcengine (Doubao) | `https://ark.cn-beijing.volces.com/api/v3` |\n| OpenAI | `https://api.openai.com/v1` |\n| SiliconFlow | `https://api.siliconflow.cn/v1` |\n| 사용자 정의 프록시 | 임의의 URL |\n\n### CLI 사용법\n\n터미널에서 직접 `review.js`를 실행할 수도 있습니다:\n\n```bash\nnode review.js --diff /tmp/my_diff.txt --model deepseek --config models.json\nnode review.js --diff /tmp/my_diff.txt --model doubao --angle \"보안 문제에 집중\" --config models.json\n```\n\n### 보안\n\n- `models.json`에는 API 키가 포함됩니다 — `.gitignore`로 제외됨\n- `review.js`는 Node.js 내장 모듈만 사용 (`npm install` 불필요)\n- API 호출은 HTTPS를 통해 이루어지며, 타임아웃은 120초\n\n---\n\n## License\n\nMIT\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fbaiyanmo%2Fclaude-code-multi-model-review","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fbaiyanmo%2Fclaude-code-multi-model-review","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fbaiyanmo%2Fclaude-code-multi-model-review/lists"}