gemini

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Gemini — Cross-Model Second Opinion

Gemini — 跨模型二次意见

You are orchestrating a cross-model review by launching Google's Gemini CLI as an independent reviewer. Gemini brings native Google ecosystem knowledge — especially valuable for Google Ads, Search Console, and SEO decisions where Google's own AI has deeper context about how their platforms work.
Unlike the code-only review pattern, this skill handles three types of changes:
  1. Code changes — diffs, new files, refactors
  2. Google Ads changes — campaign structure, bid strategies, keyword lists, negative keywords, ad copy, budget allocation
  3. SEO metadata changes — title tags, meta descriptions, schema markup, robots directives, sitemap updates, content rewrites

您将通过启动谷歌的Gemini CLI作为独立审核方,协调跨模型审核。Gemini具备原生谷歌生态系统知识——尤其在Google Ads、Search Console和SEO决策方面价值显著,因为谷歌自家AI对其平台的运作方式有更深入的了解。
与仅针对代码的审核模式不同,此技能可处理三类变更:
  1. 代码变更 — 差异对比、新文件、重构
  2. Google Ads变更 — 广告系列结构、出价策略、关键词列表、否定关键词、广告文案、预算分配
  3. SEO元数据变更 — 标题标签、元描述、Schema标记、robots指令、站点地图更新、内容重写

Step 0 — Detect Gemini CLI

步骤0 — 检测Gemini CLI

bash
command -v gemini >/dev/null 2>&1 && echo "GEMINI_FOUND" || echo "GEMINI_NOT_FOUND"
If
GEMINI_NOT_FOUND
:
Stop and tell the user:
Gemini CLI is not installed. Install it with:
npm install -g @google/gemini-cli
Then run
gemini
once to authenticate with your Google account, and retry.
If
GEMINI_FOUND
:
continue silently.

bash
command -v gemini >/dev/null 2>&1 && echo "GEMINI_FOUND" || echo "GEMINI_NOT_FOUND"
如果返回
GEMINI_NOT_FOUND
停止操作并告知用户:
Gemini CLI未安装。请使用以下命令安装:
npm install -g @google/gemini-cli
然后运行一次
gemini
以使用您的谷歌账号进行身份验证,之后重试。
如果返回
GEMINI_FOUND
继续静默执行。

Step 1 — Detect Mode

步骤1 — 检测模式

Parse the user's request to determine the mode. Match against these patterns:
ModeTrigger phrases
review"review", "check", "look at", "pass/fail", "gate", "approve"
challenge"challenge", "stress test", "break", "adversarial", "find holes", "poke holes"
consult"consult", "ask", "what does gemini think", "opinion", "advice", "strategy"
If ambiguous: default to review for changes that exist in the diff, or consult if the user is asking a question with no pending changes.

解析用户请求以确定模式。匹配以下模式:
模式触发短语
审核(review)"review"、"check"、"look at"、"pass/fail"、"gate"、"approve"
挑战(challenge)"challenge"、"stress test"、"break"、"adversarial"、"find holes"、"poke holes"
咨询(consult)"consult"、"ask"、"what does gemini think"、"opinion"、"advice"、"strategy"
如果存在歧义: 对于差异中已有的变更,默认使用审核模式;如果用户提出无待处理变更的问题,则默认使用咨询模式。

Step 2 — Detect Change Type

步骤2 — 检测变更类型

Determine what kind of changes are being reviewed. Check in this order:
确定正在审核的变更类型。按以下顺序检查:

2a — Check for Google Ads changes

2a — 检查Google Ads变更

Look for signs of Ads-related work in the current conversation context:
  • Recent MCP tool calls to
    mcp__adsagent__*
    or
    mcp__google_ads_mcp__*
  • Discussion of campaigns, keywords, bids, budgets, ad copy, negative keywords
  • Files like
    .adsagent/change-log.json
    or Ads-related config changes
If found, set
CHANGE_TYPE=google-ads
.
在当前对话语境中查找与Ads相关工作的迹象:
  • 最近对
    mcp__adsagent__*
    mcp__google_ads_mcp__*
    的MCP工具调用
  • 关于广告系列、关键词、出价、预算、广告文案、否定关键词的讨论
  • 类似
    .adsagent/change-log.json
    的文件或与Ads相关的配置变更
如果找到,设置
CHANGE_TYPE=google-ads

2b — Check for SEO metadata changes

2b — 检查SEO元数据变更

Look for:
  • Recent calls to SEO skills (seo-analysis, meta-tags-optimizer, schema-markup-generator)
  • Discussion of title tags, meta descriptions, schema markup, robots.txt, sitemaps
  • Content rewrites or keyword targeting changes
  • CMS content updates (Strapi, WordPress, etc.)
If found, set
CHANGE_TYPE=seo
.
查找:
  • 最近对SEO技能的调用(seo-analysis、meta-tags-optimizer、schema-markup-generator)
  • 关于标题标签、元描述、Schema标记、robots.txt、站点地图的讨论
  • 内容重写或关键词定位变更
  • CMS内容更新(Strapi、WordPress等)
如果找到,设置
CHANGE_TYPE=seo

2c — Check for code changes

2c — 检查代码变更

bash
git diff --stat HEAD 2>/dev/null || echo "NO_GIT_DIFF"
If there's a diff, set
CHANGE_TYPE=code
.
bash
git diff --stat HEAD 2>/dev/null || echo "NO_GIT_DIFF"
如果存在差异,设置
CHANGE_TYPE=code

2d — Mixed or unclear

2d — 混合或不明确

If multiple types are present, set
CHANGE_TYPE=mixed
. If nothing is found and mode is consult, set
CHANGE_TYPE=consult-only
.

如果存在多种类型,设置
CHANGE_TYPE=mixed
。如果未找到任何内容且模式为咨询,则设置
CHANGE_TYPE=consult-only

Step 3 — Build the Context

步骤3 — 构建上下文

Assemble the context payload that Gemini will review. Tailor it to the change type.
组装Gemini将审核的上下文负载,根据变更类型进行调整。

For
google-ads
changes:

针对
google-ads
变更:

Summarize the proposed Ads changes in a structured block:
GOOGLE ADS CHANGE SUMMARY
==========================
Account: [account name/ID if known]
Change type: [campaign creation | bid adjustment | keyword changes | negative keywords | ad copy | budget | targeting | etc.]

BEFORE (current state):
[Describe current campaign/keyword/bid state]

AFTER (proposed changes):
[Describe what will change]

BUSINESS CONTEXT:
[Goal of the change — CPA target, ROAS goal, traffic objective, etc.]
以结构化区块总结拟议的Ads变更:
GOOGLE ADS CHANGE SUMMARY
==========================
账号:[已知的账号名称/ID]
变更类型:[广告系列创建 | 出价调整 | 关键词变更 | 否定关键词 | 广告文案 | 预算 | 定位 | 等]

变更前(当前状态):
[描述当前广告系列/关键词/出价状态]

变更后(拟议变更):
[描述将要发生的变更]

业务背景:
[变更目标 — CPA目标、ROAS目标、流量目标等]

For
seo
changes:

针对
seo
变更:

SEO CHANGE SUMMARY
==================
Site: [URL]
Change type: [title tags | meta descriptions | schema markup | content rewrite | robots.txt | sitemap | etc.]

BEFORE (current state):
[Current metadata/content]

AFTER (proposed changes):
[New metadata/content]

TARGET KEYWORDS:
[Keywords being targeted, if applicable]

SEARCH INTENT:
[Informational / navigational / commercial / transactional]
SEO CHANGE SUMMARY
==================
站点:[URL]
变更类型:[标题标签 | 元描述 | Schema标记 | 内容重写 | robots.txt | 站点地图 | 等]

变更前(当前状态):
[当前元数据/内容]

变更后(拟议变更):
[新的元数据/内容]

目标关键词:
[目标关键词(如有)]

搜索意图:
[信息型 / 导航型 / 商业型 / 交易型]

For
code
changes:

针对
code
变更:

bash
BRANCH=$(git rev-parse --abbrev-ref HEAD 2>/dev/null || echo "unknown")
DIFF=$(git diff HEAD 2>/dev/null)
STAT=$(git diff --stat HEAD 2>/dev/null)
Combine the diff stat and full diff into the context.
bash
BRANCH=$(git rev-parse --abbrev-ref HEAD 2>/dev/null || echo "unknown")
DIFF=$(git diff HEAD 2>/dev/null)
STAT=$(git diff --stat HEAD 2>/dev/null)
将差异统计信息和完整差异合并到上下文中。

For
mixed
changes:

针对
mixed
变更:

Combine all applicable sections above.

合并上述所有适用部分。

Step 4 — Run Gemini

步骤4 — 运行Gemini

Build and execute the Gemini CLI command based on mode and change type.
根据模式和变更类型构建并执行Gemini CLI命令。

Review Mode

审核模式

bash
gemini -p "You are a senior reviewer with deep expertise in Google's advertising platform, Google Search, and SEO best practices. You are reviewing proposed changes for correctness, effectiveness, and potential risks.

CHANGE TYPE: ${CHANGE_TYPE}

${CONTEXT}

Evaluate these changes and produce a structured review:

1. VERDICT: PASS or FAIL (use FAIL if any blocking issue exists)

2. BLOCKING ISSUES (if any):
   - Issue, why it matters, and how to fix it

3. WARNINGS (non-blocking but worth considering):
   - Concern and recommendation

4. STRENGTHS:
   - What the changes do well

For Google Ads changes, specifically check:
- Policy compliance (disapprovals, trademark issues, restricted content)
- Budget efficiency (is spend allocated to highest-intent keywords?)
- Keyword conflicts (cannibalization, broad match pitfalls, missing negatives)
- Landing page alignment (do ads match what the page delivers?)
- Bid strategy fit (does the strategy match the campaign goal?)

For SEO changes, specifically check:
- Title tag length (under 60 chars) and keyword placement (front-loaded?)
- Meta description length (under 160 chars) and call-to-action presence
- Schema markup validity and completeness
- Potential keyword cannibalization across pages
- Search intent alignment (does the content match what users expect?)
- E-E-A-T signals (expertise, experience, authoritativeness, trustworthiness)
- Internal linking opportunities missed

For code changes, check:
- Correctness and edge cases
- Security issues
- Performance concerns
- Breaking changes" 2>&1
Capture the output. If the exit code is non-zero, report the error to the user and suggest checking
gemini
authentication.
bash
gemini -p "You are a senior reviewer with deep expertise in Google's advertising platform, Google Search, and SEO best practices. You are reviewing proposed changes for correctness, effectiveness, and potential risks.

CHANGE TYPE: ${CHANGE_TYPE}

${CONTEXT}

Evaluate these changes and produce a structured review:

1. VERDICT: PASS or FAIL (use FAIL if any blocking issue exists)

2. BLOCKING ISSUES (if any):
   - Issue, why it matters, and how to fix it

3. WARNINGS (non-blocking but worth considering):
   - Concern and recommendation

4. STRENGTHS:
   - What the changes do well

For Google Ads changes, specifically check:
- Policy compliance (disapprovals, trademark issues, restricted content)
- Budget efficiency (is spend allocated to highest-intent keywords?)
- Keyword conflicts (cannibalization, broad match pitfalls, missing negatives)
- Landing page alignment (do ads match what the page delivers?)
- Bid strategy fit (does the strategy match the campaign goal?)

For SEO changes, specifically check:
- Title tag length (under 60 chars) and keyword placement (front-loaded?)
- Meta description length (under 160 chars) and call-to-action presence
- Schema markup validity and completeness
- Potential keyword cannibalization across pages
- Search intent alignment (does the content match what users expect?)
- E-E-A-T signals (expertise, experience, authoritativeness, trustworthiness)
- Internal linking opportunities missed

For code changes, check:
- Correctness and edge cases
- Security issues
- Performance concerns
- Breaking changes" 2>&1
捕获输出。如果退出码非零,向用户报告错误并建议检查
gemini
身份验证。

Challenge Mode

挑战模式

bash
gemini -p "You are a seasoned and skeptical growth advisor who has managed eight-figure Google Ads budgets and scaled organic traffic for major brands. You have expert-level knowledge of Google's latest policies — Ads editorial standards, Performance Max behavior, broad match changes, Search quality guidelines, spam policies, Core Web Vitals thresholds, and structured data requirements.

Your role is devil's advocate. The team is proposing changes and they want you to pressure-test them before committing. Do not be agreeable — your value is in catching what optimism misses. Evaluate based on evidence, data, and your experience with how Google's systems actually behave (not how documentation says they should).

CHANGE TYPE: ${CHANGE_TYPE}

${CONTEXT}

For each proposed change:

1. STATE THE ASSUMPTION — What is the team assuming will happen?
2. CHALLENGE IT — Why might that assumption be wrong? Cite specific Google policy, algorithm behavior, or auction mechanics where relevant. Reference real patterns you'd expect to see in the data.
3. WHAT DOES THE DATA SAY? — What metrics or signals should the team check before and after to validate this change? Be specific (e.g. 'compare impression share lost to rank before and 14 days after', not 'monitor performance').
4. VERDICT — For each change: SOUND, RISKY, or RETHINK. One sentence explaining why.

Finally, give an overall honest opinion: is this set of changes worth shipping as-is, or should the team pause and address specific concerns first? Be concise and professional — no filler, no hedging." 2>&1
bash
gemini -p "You are a seasoned and skeptical growth advisor who has managed eight-figure Google Ads budgets and scaled organic traffic for major brands. You have expert-level knowledge of Google's latest policies — Ads editorial standards, Performance Max behavior, broad match changes, Search quality guidelines, spam policies, Core Web Vitals thresholds, and structured data requirements.

Your role is devil's advocate. The team is proposing changes and they want you to pressure-test them before committing. Do not be agreeable — your value is in catching what optimism misses. Evaluate based on evidence, data, and your experience with how Google's systems actually behave (not how documentation says they should).

CHANGE TYPE: ${CHANGE_TYPE}

${CONTEXT}

For each proposed change:

1. STATE THE ASSUMPTION — What is the team assuming will happen?
2. CHALLENGE IT — Why might that assumption be wrong? Cite specific Google policy, algorithm behavior, or auction mechanics where relevant. Reference real patterns you'd expect to see in the data.
3. WHAT DOES THE DATA SAY? — What metrics or signals should the team check before and after to validate this change? Be specific (e.g. 'compare impression share lost to rank before and 14 days after', not 'monitor performance').
4. VERDICT — For each change: SOUND, RISKY, or RETHINK. One sentence explaining why.

Finally, give an overall honest opinion: is this set of changes worth shipping as-is, or should the team pause and address specific concerns first? Be concise and professional — no filler, no hedging." 2>&1

Consult Mode

咨询模式

bash
gemini -p "You are a Google Ads and SEO expert consultant with deep knowledge of Google's ecosystem — Search algorithms, Ads auction mechanics, Search Console, and web performance. The user wants your independent perspective.

CONTEXT:
${CONTEXT}

USER QUESTION:
${USER_QUESTION}

Provide a clear, opinionated answer. If you disagree with a proposed approach, say so directly and explain why. Draw on Google-specific knowledge — ad auction dynamics, search ranking factors, Quality Score mechanics, Core Web Vitals thresholds, etc." 2>&1

bash
gemini -p "You are a Google Ads and SEO expert consultant with deep knowledge of Google's ecosystem — Search algorithms, Ads auction mechanics, Search Console, and web performance. The user wants your independent perspective.

CONTEXT:
${CONTEXT}

USER QUESTION:
${USER_QUESTION}

Provide a clear, opinionated answer. If you disagree with a proposed approach, say so directly and explain why. Draw on Google-specific knowledge — ad auction dynamics, search ranking factors, Quality Score mechanics, Core Web Vitals thresholds, etc." 2>&1

Step 5 — Present Results

步骤5 — 呈现结果

5a — Format the Gemini output

5a — 格式化Gemini输出

Present Gemini's response with a clear header:
Gemini Review (
${CHANGE_TYPE}
|
${MODE}
mode)
[Gemini's formatted output]
使用清晰的标题呈现Gemini的响应:
Gemini审核 (
${CHANGE_TYPE}
|
${MODE}
模式)
[Gemini的格式化输出]

5b — Cross-model analysis (if Claude already reviewed)

5b — 跨模型分析(如果Claude已审核)

If Claude has already reviewed the same changes (e.g., via
/toprank:seo-analysis
or
/toprank:ads-audit
earlier in the conversation), produce a cross-model comparison:
Cross-Model Analysis: Claude vs Gemini
Overlapping findings (both flagged):
  • [Finding 1]
  • [Finding 2]
Claude-only findings:
  • [Finding that only Claude caught]
Gemini-only findings:
  • [Finding that only Gemini caught]
Disagreements (if any):
  • [Topic]: Claude says X, Gemini says Y
Overlapping findings have higher confidence — they should be addressed first. Unique findings from either model are worth investigating. Disagreements should be flagged to the user for a judgment call.
如果Claude已审核相同的变更(例如,对话中较早使用
/toprank:seo-analysis
/toprank:ads-audit
),生成跨模型对比:
跨模型分析:Claude vs Gemini
重叠发现(两者均标记):
  • [发现1]
  • [发现2]
仅Claude发现的问题:
  • [仅Claude捕获的发现]
仅Gemini发现的问题:
  • [仅Gemini捕获的发现]
分歧点(如有):
  • [主题]:Claude认为X,Gemini认为Y
重叠发现的可信度更高——应优先处理。任一模型的独特发现都值得调查。分歧点应标记给用户以进行判断。

5c — Suggest next steps

5c — 建议下一步行动

Based on the results:
  • Review PASS: "Gemini approved. Ready to ship."
  • Review FAIL: "Gemini flagged blocking issues. Address them, then re-run
    /toprank:gemini review
    ."
  • Challenge — HIGH risk: "Stress test surfaced high-risk scenarios. Consider the mitigations before proceeding."
  • Challenge — LOW risk: "Gemini couldn't find major attack vectors. Changes look resilient."
  • Consult: "Want me to apply any of Gemini's suggestions?"
根据结果:
  • 审核通过(PASS): "Gemini已批准。可发布。"
  • 审核失败(FAIL): "Gemini标记了阻塞性问题。解决后重新运行
    /toprank:gemini review
    。"
  • 挑战模式——高风险: "压力测试发现高风险场景。请在推进前考虑缓解措施。"
  • 挑战模式——低风险: "Gemini未发现主要漏洞。变更看起来具备韧性。"
  • 咨询模式: "需要我应用Gemini的任何建议吗?"