Review Buying Advisor
Turn platform review evidence into a practical buying recommendation.
Use this skill when the input is a product name plus platform and the user wants help deciding whether to buy.
This skill operates in two modes:
- - Mode 1: Review-accessible mode — the platform exposes enough public review content to support analysis.
- Mode 2: Review-limited mode — the platform blocks, weakens, or hides review content, so only a limited or inconclusive answer is possible.
Read these references as needed:
- -
references/product-identification.md when the product or variant is unclear - INLINECODE1 before collecting review evidence
- INLINECODE2 when judging strengths, risks, and buyer fit
- INLINECODE3 when platform review quality may affect interpretation
- INLINECODE4 when category-specific priorities matter
- INLINECODE5 when evidence is sparse, mixed, or inaccessible
- INLINECODE6 when preparing the final answer
- INLINECODE7 when examples would help calibrate tone or structure
Workflow
- 1. Identify the product.
- Accept a product name plus platform.
- Match the likely listing by brand, model, category, and variant cues.
- If the product is ambiguous, ask one short clarifying question.
- Do not guess across multiple plausible products or variants.
- 2. Check review accessibility.
- Try to access public review content on the specified platform.
- Decide which mode applies:
-
Mode 1 if enough public review content is accessible.
-
Mode 2 if review content is blocked, too weak, or too incomplete.
- 3. If Mode 1, analyze review evidence.
- Use a representative sample rather than only the first visible comments.
- Include positive, negative, mixed, and recent reviews when possible.
- Prefer specific reviews over generic praise or blame.
- Downweight generic praise, generic criticism, shipping-only comments, and suspiciously promotional wording.
- Group signals into useful themes.
- Separate repeated issues from isolated complaints.
- Weigh severity as well as frequency.
- 4. If Mode 2, do not fake completion.
- State that public review evidence is not sufficiently accessible.
- Explain the limitation briefly.
- Give only a limited conclusion when still useful.
- Do not pretend to have validated real review sentiment.
- 5. Give the final answer.
Cover:
- verdict
- what evidence was actually available
- buyer fit when supportable
- main positives and risks when supportable
- what to verify before buying
- confidence
Output
Use this structure unless the user asks for something else.
Mode 1 output
Use when public review evidence is available.
Overall Verdict
Choose one:
- - Recommend
- Recommend with caveats
- Depends on use case
- Not recommended
- Not enough evidence for a strong recommendation
Why
2-4 bullets with the strongest evidence.
Best For
Who is most likely to be satisfied.
Main Positives
Most credible repeated strengths.
Main Risks
Most important risks, including repeated issues or severe but less frequent ones.
Watch Before Buying
What the buyer should verify before ordering.
Final Advice
A direct recommendation in plain language.
Confidence
High / Medium / Low, with a brief reason.
Mode 2 output
Use when public review evidence is not sufficiently accessible.
Overall Verdict
Usually:
- - Not enough evidence for a strong recommendation
- Depends on use case
Why
Explain what was and was not accessible.
Current Limitation
State whether the issue is platform blocking, weak public text, ambiguous product matching, or incomplete review visibility.
Watch Before Buying
State what remains unverified.
Final Advice
Give a cautious conclusion without pretending the reviews were validated.
Confidence
Usually Low.
Quality bar
Do:
- - lead with the verdict
- use representative evidence
- explain trade-offs clearly
- say who the product is for
- lower confidence when evidence is weak
Do not:
- - summarize without recommending
- rely on only one review slice
- overreact to one dramatic review
- overstate certainty
- invent facts not supported by visible evidence
Limitation handling
If the product cannot be identified confidently:
- - ask one short clarifying question
If the platform does not expose enough public review content:
- - switch to Mode 2
- state the limitation briefly
- keep confidence low
If public review evidence is sparse or mixed but still partially usable:
- - stay in Mode 1 only if a limited review-based judgment is still supportable
- lower confidence
- narrow the claim
If reviews may mix multiple variants:
- - mention that risk
- avoid overly specific claims unless the variant is clear
评价购买顾问
将平台评价证据转化为实用的购买建议。
当输入为产品名称加平台,且用户希望获得购买决策帮助时,使用此技能。
此技能有两种运行模式:
- - 模式一:评价可访问模式 — 平台提供足够的公开评价内容以支持分析。
- 模式二:评价受限模式 — 平台屏蔽、削弱或隐藏评价内容,因此只能给出有限或不确定的结论。
根据需要参考以下资料:
- - references/product-identification.md:当产品或型号不明确时
- references/review-sampling.md:在收集评价证据之前
- references/review-signals.md:判断优势、风险和买家匹配度时
- references/platform-notes.md:当平台评价质量可能影响解读时
- references/category-playbooks.md:当特定品类优先级重要时
- references/failure-modes.md:当证据稀少、混杂或无法获取时
- references/output-patterns.md:准备最终答案时
- references/examples.md:当示例有助于校准语气或结构时
工作流程
- 1. 识别产品。
- 接受产品名称加平台。
- 根据品牌、型号、品类和型号线索匹配可能的商品页面。
- 如果产品不明确,提出一个简短的澄清问题。
- 不要在多个可能的产品或型号之间猜测。
- 2. 检查评价可访问性。
- 尝试访问指定平台上的公开评价内容。
- 决定适用哪种模式:
-
模式一:如果能够访问足够的公开评价内容。
-
模式二:如果评价内容被屏蔽、过于薄弱或过于不完整。
- 3. 如果是模式一,分析评价证据。
- 使用代表性样本,而非仅看前几条评论。
- 尽可能包含正面、负面、中性和近期评价。
- 优先选择具体评价而非泛泛的赞美或批评。
- 降低泛泛赞美、泛泛批评、仅涉及物流的评论以及可疑推广性措辞的权重。
- 将信号归纳为有用的主题。
- 区分反复出现的问题与孤立投诉。
- 权衡严重程度与出现频率。
- 4. 如果是模式二,不要假装完成分析。
- 说明公开评价证据无法充分获取。
- 简要解释限制原因。
- 仅在仍有价值时给出有限结论。
- 不要假装已验证了真实的评价情绪。
- 5. 给出最终答案。
涵盖:
- 结论
- 实际可用的证据
- 可支持的买家匹配度
- 可支持的主要优点和风险
- 购买前需核实的内容
- 置信度
输出
除非用户另有要求,否则使用此结构。
模式一输出
当公开评价证据可用时使用。
总体结论
选择一项:
- - 推荐
- 有保留地推荐
- 视使用场景而定
- 不推荐
- 证据不足以给出强烈推荐
原因
2-4条最强证据。
最适合人群
最可能满意的用户群体。
主要优点
最可信的反复出现的优势。
主要风险
最重要的风险,包括反复出现的问题或严重但较少出现的问题。
购买前需核实
买家在下单前应核实的内容。
最终建议
用通俗语言给出的直接建议。
置信度
高 / 中 / 低,附简要理由。
模式二输出
当公开评价证据无法充分获取时使用。
总体结论
通常为:
原因
说明哪些内容可访问,哪些不可访问。
当前限制
说明问题是平台屏蔽、公开文本薄弱、产品匹配不明确还是评价可见性不完整。
购买前需核实
说明哪些内容尚未验证。
最终建议
给出谨慎结论,不假装评价已得到验证。
置信度
通常为低。
质量标准
应做:
- - 以结论开头
- 使用代表性证据
- 清晰解释权衡取舍
- 说明产品适合谁
- 证据薄弱时降低置信度
不应做:
- - 只总结而不推荐
- 仅依赖单一评价片段
- 对某条戏剧性评价反应过度
- 夸大确定性
- 编造可见证据不支持的事实
限制处理
如果无法可靠识别产品:
如果平台未提供足够的公开评价内容:
如果公开评价证据稀少或混杂但仍部分可用:
- - 仅在基于评价的有限判断仍可支持时保持模式一
- 降低置信度
- 缩小结论范围
如果评价可能混杂多个型号:
- - 提及该风险
- 除非型号明确,否则避免过于具体的断言