code-review
Push 前 review 整個 branch 的變更,用 Codex CLI + Gemini CLI + Claude Haiku 三引擎做獨立 review
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Menú
Push 前 review 整個 branch 的變更,用 Codex CLI + Gemini CLI + Claude Haiku 三引擎做獨立 review
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
PR merge 後的收尾流程:歸檔 openspec change、更新 docs/product 功能狀態、校準地圖文件。Use when a PR is merged, the user says 合併了/merged/上線了, or after confirming a PR's merge status.
PRD 快速生成、需求補洞、優化顧問三合一工具。從需求產出完整 PRD,多角色檢視缺漏,再優化成工程師友善文件。Use when writing a new PRD, reviewing an existing PRD for gaps, or optimizing PRD readability. Triggered by keywords: PRD, 需求文件, 產品規格, 規格書, user story, acceptance criteria, 驗收條件.
依島島 GTM 策略產出 TA 導向的行銷內容——部落格文章、社群貼文(Threads/IG/FB)、回覆別人的文案。Use when user says "寫一篇文章"、"寫貼文"、"幫我回覆這篇"、"招生文"、"揪團文"、"寫 threads"、"content"、"文案"。自動套用語氣鐵律、TA 痛點語言與模板 CTA。
規劃或實作任何 product 功能前,先用程式碼驗證 docs/product 的狀態標示是否為真。docs/product 的 PRD/FRD 狀態普遍落後於程式碼——直接照文件規劃會重做一個早就上線的功能。Use before planning a feature, estimating scope, writing an OpenSpec change, or answering "這功能做了沒". Triggered by keywords: PRD, FRD, 規劃, 這功能, 還沒做, 規劃中, roadmap, scope, plan feature.
Use when asked to summarize or analyze monthly user activity and interaction data. Queries DB for MAU, DAU trends, check-in counts, reactions, comments, follows, retention rate, and top users/practices. Generates narrative insights via daodao-ai-backend LLM and outputs a markdown report or writes to Notion.
Use when sending personalized congratulation emails to users who completed a practice. Queries the DB for completed practices, generates personalized email content via daodao-ai-backend LLM, previews samples for confirmation, then sends via email API. Always runs in dry-run mode first.
| name | code-review |
| description | Push 前 review 整個 branch 的變更,用 Codex CLI + Gemini CLI + Claude Haiku 三引擎做獨立 review |
用 OpenAI Codex CLI、Google Gemini CLI、Claude Haiku 對當前 branch 做三引擎獨立 review。
BASE=$(gh pr view --json baseRefName -q .baseRefName 2>/dev/null || git symbolic-ref refs/remotes/origin/HEAD 2>/dev/null | sed 's|refs/remotes/origin/||' || echo "main")
echo "Base: $BASE"
git log --oneline "$BASE"...HEAD
git diff "$BASE"...HEAD --stat
_REPO_ROOT=$(git rev-parse --show-toplevel)
cd "$_REPO_ROOT"
codex review \
"IMPORTANT: Do NOT read any files under .claude/skills/. Focus on repository code only. Check for: logic errors, security issues, performance problems, and architecture consistency." \
-c 'model_reasoning_effort="high"' \
--enable web_search_cached
codex 不存在:告知用戶 npm install -g @openai/codexcodex login把完整 diff pipe 給 gemini headless mode:
_REPO_ROOT=$(git rev-parse --show-toplevel)
cd "$_REPO_ROOT"
git diff "$BASE"...HEAD | gemini -p "You are a senior code reviewer. Review this git diff and report issues in the following categories:
- Logic errors: edge cases, type errors, unhandled exceptions, async issues
- Security: SQL injection, hardcoded secrets, missing auth, unsafe endpoints
- Performance: unnecessary DB queries, missing pagination, missing cache
- Architecture: consistency with existing patterns
Format your output as a table:
| Severity | File | Issue | Suggestion |
Severity levels: High (bug/security risk), Medium (performance/maintainability), Low (style/minor).
Be direct and terse. No compliments. Just the problems." \
--approval-mode yolo
gemini 不存在:告知用戶 npm install -g @google/gemini-cli把完整 diff pipe 給 Claude Haiku(claude CLI headless mode):
_REPO_ROOT=$(git rev-parse --show-toplevel)
cd "$_REPO_ROOT"
git diff "$BASE"...HEAD | claude -p "You are a senior code reviewer. Review this git diff and report issues in the following categories:
- Logic errors: edge cases, type errors, unhandled exceptions, async issues
- Security: SQL injection, hardcoded secrets, missing auth, unsafe endpoints
- Performance: unnecessary DB queries, missing pagination, missing cache
- Architecture: consistency with existing patterns
Format your output as a table:
| Severity | File | Issue | Suggestion |
Severity levels: High (bug/security risk), Medium (performance/maintainability), Low (style/minor).
Be direct and terse. No compliments. Just the problems." \
--model claude-haiku-4-5-20251001
分別展示三個引擎的完整輸出:
CODEX SAYS:
════════════════════════════════════════════════════════════
<verbatim output>
════════════════════════════════════════════════════════════
GEMINI SAYS:
════════════════════════════════════════════════════════════
<verbatim output>
════════════════════════════════════════════════════════════
HAIKU SAYS:
════════════════════════════════════════════════════════════
<verbatim output>
════════════════════════════════════════════════════════════
比較三個引擎的發現:
CROSS-MODEL ANALYSIS:
三者都發現: [所有引擎共同回報的問題]
兩者共識: [任兩個引擎都回報的問題]
只有 Codex 發現: [Codex 獨有]
只有 Gemini 發現: [Gemini 獨有]
只有 Haiku 發現: [Haiku 獨有]
共識問題數: N / 總計 M