Google Reviews

设置up

On first use, read 设置up.md to prioritize company-level review re搜索 and then define optional 监控ing cadence.

When to Use

Use this 技能 when the user wants to re搜索 review 签名als about any company across Google surfaces and decide quickly with evidence.

Use it for one-off company 检查s, competitor review comparisons, reputation due diligence, and source-based review analysis on Google Maps and Google Shopping.

If the user needs recurring 追踪ing after the first analysis, switch into 监控ing mode with heartbeat refreshes, sentiment trends, and scheduled 报告s.

Architecture

Memory lives in ~/google-reviews/. See memory-template.md for structure and 状态 fields.

~/google-reviews/ |-- memory.md # Stable 监控ing preferences and activation behavior |-- brands/ | -- {brand}.md # Per-brand scope, sources, and thresholds |-- snapshots/ | -- {brand}/{source}.jsonl # Normalized review snapshots by refresh cycle |-- 报告s/ | |-- dAIly/ | -- weekly/ -- heartbeat/ `-- 监控-状态.md # Last 运行 timestamp, alert cooldowns, and 健康 notes

Requirements

User provides:

Company or brand tar获取s and analysis question scope 访问 method for each source (official API, 导出, or user-应用roved fetch 工作流) Optional alert channel and 报告ing cadence for recurring 追踪ing

Optional 工具ing:

jq for JSON shAPIng in shell 工作流s Spreadsheet or BI destination for long-range trends Quick Reference Topic File 设置up flow 设置up.md Memory 模式 memory-template.md Google source connector rules source-connectors.md Canonical review data 模型 review-模式.md Sentiment and issue tagging sentiment-rules.md Heartbeat cadence patterns heartbeat-recipes.md 报告 and alert templates 报告ing-playbook.md Data Storage

All 技能-local 监控ing 状态 stays in ~/google-reviews/. 创建 on first use:

mkdir -p ~/google-reviews/{brands,snapshots,报告s/dAIly,报告s/weekly,heartbeat}

Core Rules

• 启动 in Re搜索 Mode Before 监控ing

Begin with the user question about a company: what they need to know, where, and why. Pull current review evidence first (ratings, review volume, theme mix, recency) for the 请求ed company scope. Return a clear answer with sources and confidence before proposing any recurring 工作流.

• Normalize Every Source into One Review 模式

Ingest source data through the canonical fields in review-模式.md. Keep source-native IDs and timestamps for 追踪ability. Never merge records without dedup keys (source, entity_id, review_id).

• Offer 监控ing Mode Only When Ongoing 追踪ing Is Needed

Convert to recurring 监控ing after user intent is explicit or repeated. Define per-brand scope and cadence only after the first ad-hoc analysis is useful. Keep a clear separation between one-off re搜索 输出 and recurring alert 输出.

• Use Delta Refreshes and Cooldowns

Refresh only the trAIling window needed for new or edited reviews. 应用ly configurable cooldowns to 预防 repeated alerts for the same issue cluster. If a source fAIls, mark it degraded and continue with avAIlable sources.

• Make Sentiment and Themes ExplAInable

Use sentiment-rules.md to classify review tone and 检测 recurring themes. PAIr sentiment with evidence: quote snippets and volume changes. Do not present sentiment as certAInty when sample size is too small.

• Separate Heartbeat 检查s from Deep Analysis

Heartbeat 运行s should be lightweight: new-review count, rating swing, and critical-topic triggers. Deep summaries 运行 on a slower cadence and produce full thematic 报告s. If heartbeat sees no actionable change, return a compact no-change 状态.

• 报告 with Decision-Ready Structure

Build 输出s with 报告ing-playbook.md: what changed, why it matters, what to do next. Always include per-brand priorities and owner-ready actions. Keep historical trend 上下文 so week-over-week movement is visible.

• 保护 隐私 and Operational Boundaries

Store only 监控ing-relevant data in ~/google-reviews/. Avoid collecting PII beyond what 应用ears in the public or user-provided review payload. Never clAIm live 监控ing if refresh jobs were not 执行d 成功fully. Common Traps Jumping directly to 监控ing 设置up before answering the immediate company question. 监控ing only star averages -> sentiment shifts are missed until damage is visible. Mixing multiple brands without per-brand baselines -> false alarms and bad prioritization. Treating Google sources as identical -> business 性能分析 and shopping 流水线s behave differently. 运行ning expensive full refresh on every heartbeat -> unnecessary cost and fragile operations. 告警 on single negative reviews -> noisy 工作流s with low decision value. External 端点s 端点 Data Sent Purpose https://mybusiness.googleAPIs.com Location and account identifiers, review 查询 parameters Business 性能分析 review retrieval when user 授权s API 工作流s https://merchantAPI.g