安装指南

1. ClawHub CLI 从源直接安装技能的最快方式。
   npx clawhub@latest install brand-mention-research
2. 手动安装 将技能文件夹复制到以下位置之一
   全局模式
   ~/.openclaw/skills/
   工作区
   /skills/
   优先级：工作区 > 本地 > 内置
3. 提示词安装 将此提示词复制到 OpenClaw 即可自动安装。
   请帮我使用 Clawhub 安装 brand-mention-research。如果尚未安装 Clawhub，请先安装（npm i -g clawhub）。
   更倾向于下载？
   获取包含原生技能文件的ZIP压缩包。
   下载技能 (.zip)

什么是

品牌提及研究借助 Crawlora，帮助你在 Reddit、TikTok 和 YouTube 上调查品牌、产品或营销活动。它会解析别名，根据来源详情核查所选结果，在各平台内去重，并将观察结果分类为已确认、模糊、推广、引用或误报。
该技能优先考虑可追溯的证据，而非夸大的触达声称：它记录搜索范围和抽样限制，区分帖子与抽样评论，并报告比例的分母。作为 Openclaw Skills 的一部分，它支持实用的声誉研究，不会把搜索结果数量当成提及量，也不会得出超出可用证据范围的结论。

应用场景

汇总所选社交平台上有关品牌或产品讨论的、有来源依据的快照。
解析品牌别名，并从常见名称的无关结果中辨别真实提及。
使用相同的日期窗口和抽样预算比较讨论主题或竞争对手。
查看相关帖子及其有限样本评论，以调查产品问题。
保存相同的查询、范围和抽样计划，开展可重复的研究。
编写声誉简报，清楚区分已确认、模糊、推广、引用和排除的观察结果。

工作原理

定义实体、别名、可选竞争对手、日期窗口、语言或市场、目标平台、页面预算和额度限制。
配置 Crawlora 访问权限，并选择可用的 Crawlora MCP 工具或捆绑的 shell 脚本。
在所请求的平台上运行小规模初始搜索，保留查询词、筛选条件、结果顺序、分页信息和检索时间。
评估相关性；不要翻阅大多无关的结果页，而应收紧范围过宽或精确度较低的搜索。
将选定的搜索结果解析为帖子或视频详情，并在每个平台内使用规范 ID 去重。
仅为选定的相关帖子获取评论，并保留父帖子链接、可用时的评论身份、顺序和样本数量。
对每条观察进行分类，谨慎解读情绪，并将未知或不确定信息与已核实事实区分开来。
提供覆盖记录表和证据表，列出来源详情、摘录、主题、情绪、选择理由、计数和分母。

配置指南

将 CRAWLORA_API_KEY 设为你的 Crawlora API 密钥。
连接后优先使用列出的 Crawlora MCP 工具。否则，使用捆绑的 scripts/crawlora.sh 脚本，该脚本会向 Crawlora API 发送 x-api-key 标头。
查阅 reference/endpoints.md，了解准确的端点方法和参数。
除 HTTP 状态外，还要检查应用层 code；响应负载位于 data 中。
遇到身份验证错误时停止；收到 HTTP 429 响应时退避；上游出现暂时性故障时重试一次。
确保调用次数和页面数不超出请求范围及额度预算。
示例搜索：
scripts/crawlora.sh /reddit/search q="Crawlora" sort=top time=month limit=25
scripts/crawlora.sh /tiktok/search keyword="Crawlora" count=10
scripts/crawlora.sh /youtube/search q="Crawlora" type=video upload_date=this_month

请求所选详情或评论前，先解析返回的 ID：

scripts/crawlora.sh "/reddit/comments/$POST_ID" limit=25

数据架构

围绕已定义的实体和可重复的搜索范围组织研究。保留所提供的来源 ID、URL、检索时间和来源日期。
记录或字段 保留内容
范围 品牌或产品、别名、按需提供的竞争对手、日期窗口、语言或市场、平台以及页面或额度预算
搜索覆盖 平台、查询或别名、筛选条件、结果顺序、页面或游标以及检索时间
分类 已确认提及、模糊匹配、推广提及、引用，或附有理由的排除误报
解读 主题和情绪；内容或意图不确定时使用混合或未知
评论 父帖子、可用时的评论身份、顺序、分页信息和抽样评论数量
报告 唯一的已确认帖子、排除或模糊匹配、抽样评论，以及所报告比例的分母
将顶层帖子数量与评论数量及样本分开。将缺失的互动指标视为未知。按原样保留日期；此工作流程中的 TikTok 结果没有日期筛选参数，因此应在本地筛选可用日期，并将未知日期标记为未知，不要虚构后端筛选器。Reddit 的 time 筛选器适用于 top 和 comments 排序；YouTube 支持文档所述的 upload_date 及区域设置或地区字段。

高级特性

别名解析会使用相互佐证的产品、域名、账号名或上下文证据，减少常见名称造成的误匹配。
平台专属搜索处理会记录查询设置，并遵守 Reddit、TikTok 和 YouTube 筛选器的文档限制。
规范帖子 ID 支持平台内去重；可将跨平台转载和引用关联为相关观察，而不假定文本相同就代表作者相同。
有选择地收集评论会保留父帖子关系，并将评论样本与顶层结果区分开来。
情绪和提及标签明确支持模糊、混合和未知情况，而非强行得出确定结论。
可重复的覆盖记录表让抽样限制和证据在不同运行中均可追溯。
竞争对手比较要求使用相同的时间窗口和预算，同时记录平台或搜索覆盖不一致之处。
除非收到请求，否则此工作流程不会提出缺乏依据的受众触达或市场份额声称，也不会联系作者、发布回复或安排监测。

name: brand-mention-research description: Build a sourced brand-mention and reputation brief from bounded Reddit, TikTok, and YouTube searches and comment samples through Crawlora. Use to resolve brand aliases, deduplicate mentions, compare discussion themes, or inspect a product issue with explicit sampling limits.

Brand mention research

Access and scope

Set CRAWLORA_API_KEY to your Crawlora key. Prefer the listed Crawlora MCP tools when connected; otherwise use the bundled scripts/crawlora.sh, which sends x-api-key to https://api.crawlora.net/api/v1. Read reference/endpoints.md for exact methods and parameters. Check application code as well as HTTP status; payloads are inside data. Stop on authentication errors, back off on 429, and retry a transient upstream failure once. Bound pages and calls by the requested scope and credit budget. Preserve source IDs, URLs, retrieval times, and source dates where supplied.

Produce a bounded mention brief for a brand, product, or campaign. Define the entity and aliases, competitors when requested, date window, languages/markets, platforms, and page budget before collecting. Common names need corroborating product, domain, handle, or contextual evidence to avoid false matches. Start with a small sample for each query. If broad Reddit or platform search results are mostly unrelated, tighten the exact terms or select a relevant community; do not treat returned-result counts as mention volume or continue paging through a low-precision query.

Collect and classify attributable mentions

Search Reddit, TikTok, or YouTube according to the requested sources. Retain query/alias, filters, result order, pages/cursors, and retrieval time. Reddit's time filter applies to top/comments sorts; do not claim sort=new time=week enforces a week. TikTok has no date-filter argument here: filter returned dates locally and mark unknown dates rather than inventing a backend filter. YouTube uses its documented upload_date and locale/region fields.
Resolve selected results into post/video detail. Canonical post IDs are the within-platform deduplication key. Cross-posted or quoted content can be linked as related observations, but identical text does not prove the same author.
Fetch comments only for selected relevant posts. Preserve parent post, comment identity when available, ordering, and pagination. Keep comment counts/sample size separate from top-level post counts. Reddit's metrics mode costs more; request it only when engagement data is needed. Missing engagement remains unknown.
Label each observation as a confirmed mention, ambiguous match, promotional mention, quotation, or excluded false positive with a reason. Interpret sentiment from available text/context; use mixed or unknown for sarcasm, missing content, or uncertain intent. Do not invent video dialogue from a caption alone.
scripts/crawlora.sh /reddit/search q="Crawlora" sort=top time=month limit=25
scripts/crawlora.sh /tiktok/search keyword="Crawlora" count=10
scripts/crawlora.sh /youtube/search q="Crawlora" type=video upload_date=this_month

# Resolve returned IDs before requesting selected details/comments:
# scripts/crawlora.sh "/reddit/comments/$POST_ID" limit=25

Deliver a reproducible brief

Return a coverage ledger and evidence table with platform, source URL/ID, publication time if available, alias/query, short excerpt, theme, sentiment, and explicit selection reason. Show unique confirmed posts, excluded/ambiguous matches, sampled comments, and denominators for each reported proportion. Compare competitors only under equivalent windows and sampling budgets, and still describe unequal platform/search coverage. Do not sum follower counts or views into unique audience reach, infer market share from mention counts, or call a sampled issue widespread without corroborating evidence. Repeated runs need the same saved query/scope to support change claims. Do not contact authors, post responses, or schedule monitoring unless the user requests those actions.

足球比赛表现分析
Caly 预约管理器