安装指南

1. ClawHub CLI 从源直接安装技能的最快方式。
   npx clawhub@latest install reddit-stock-sentiment-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 reddit-stock-sentiment-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

Reddit 股票情绪研究可帮助 AI 代理调查人们对上市公司或股票代码的讨论。它会核实证券信息，搜索相关 Reddit 讨论，将相关结果与含糊或无关的匹配项区分开来，并按立场和主题对抽样帖子与评论进行分类。此工作流程旨在呈现证据和覆盖范围限制，而不是将样本视为所有投资者的代表。
在有帮助时，该技能会为同一时间窗口添加 Yahoo Finance 报价或历史价格背景。它将观察到的市场走势与 Reddit 用户观点分开，避免把传闻当作事实，也不会给出价格预测或买入、卖出、持有建议。在 Openclaw Skills 中，它适用于对公开市场讨论开展结构化、带来源链接的研究。

应用场景

回答“Reddit 上的投资者对 NVDA 有何看法？”之类的问题，并提供链接示例和明确的样本限制。
比较选定 Reddit 主题帖中对某公司的正面、负面、褒贬不一、中立或不明确的观点。
识别产品体验、财务表现、估值、催化因素、风险或投机等讨论主题。
添加与指定讨论时间窗口相匹配的 Yahoo Finance 历史价格背景。
调查可能是普通词语或其他公司代码的股票代码，先核实发行方和市场背景。
生成关注来源的研究简报，而不把抽样的社交讨论转化为交易信号或预测。

工作原理

确认股票代码、发行方、交易所或市场、请求的时间窗口、相关社区和研究问题。通过 Yahoo Finance 搜索解决含糊的代码，并核实选定的证券。
使用经核实的股票代码和发行方名称作为不同查询变体搜索 Reddit。记录查询、社区筛选条件、排序和时间设置；检查少量初始结果，并将候选结果标记为相关、含糊或无关。
通过优化嘈杂的查询或 subreddit 筛选条件来减少噪声，而不是深入翻页查找误匹配。仅为选定的相关主题帖获取原始帖子和评论，按 ID 对帖子去重，并分别记录帖子和评论样本数量。
必要时在本地检查来源日期。将未知日期和缺失的互动字段视为未知，而不是零；请记住，Reddit 的时间筛选对于 new 排序并不是可靠的日期筛选方式。
如果价格背景有用，则解析股票代码并获取讨论时间窗口内的报价或历史序列。记录币种、间隔、准确日期以及价格是否经过调整；将市场交易时段与帖子日期对齐。
对实际表达的观点进行分类，并标注讨论主题。链接具有代表性的示例，纳入反例，并避免推断评论者身份、持仓或交易活动。
报告证券及研究范围、检索时间、查询详情、相关和排除数量、样本量、按主题整理的证据、可选的价格背景以及覆盖范围限制。不要声称样本具有代表性，也不要建议进行交易。

配置指南

使用随附的 scripts/crawlora.sh 辅助脚本。将 Crawlora API 密钥配置为 CRAWLORA_API_KEY；辅助脚本会通过 HTTP 标头发送该密钥。具体路由和参数请参阅 reference/endpoints.md。

解析并核实公司或股票代码

scripts/crawlora.sh /yahoo-finance/search q="NVIDIA" quotes_count=5

使用股票代码和发行方名称变体搜索 Reddit

scripts/crawlora.sh /reddit/search q="NVDA" sort=top time=month limit=10
scripts/crawlora.sh /reddit/search q="NVIDIA" subreddit=stocks sort=top time=month limit=10

选定相关帖子并确认证券信息后，获取评论

scripts/crawlora.sh /reddit/comments/
sort=top limit=25

可选：获取匹配时间窗口的历史价格数据

scripts/crawlora.sh /yahoo-finance/ticker//history period=1mo interval=1d
在示例命令中替换为经核实的帖子 ID 和股票代码。选择适合研究问题的时间窗口和 subreddit 范围，默认使用普通的公开 Reddit 信息流。仅在明确需要互动数据时才请求成本较高的 metrics 模式。
Reddit 股票情绪研究 数据架构与分类体系
证券身份： 经核实的股票代码、发行方、交易所或市场，以及（使用价格数据时）币种。
研究范围： 讨论时间窗口、检索时间、Reddit 查询、subreddit 筛选条件、排序和时间设置。
候选项分类： 每个经过审核的结果都标记为相关、含糊或无关；选定的帖子按帖子 ID 去重。
样本计数： 相关、含糊和排除的帖子数量分别报告，并与主题帖和抽样评论数量分开。一个主题帖中的重复评论并不代表相互独立的投资者观点。
证据标签： 立场分为正面、负面、褒贬不一、中立/提问或不明确。另可使用产品体验、财务表现、估值、催化因素、风险或投机等主题标签。
证据引用： 具有代表性的帖子链接和反例用于支持主题总结；谨慎少量引用原文。
可选价格背景： Yahoo Finance 股票代码、币种、间隔、准确的开始和结束日期以及调整依据。
覆盖范围元数据： 明确说明未知日期、无法获取的互动字段、检索限制和未解答的问题。若没有明确的抽样分母，则不报告总体情绪百分比。

高级特性

含糊股票代码处理：
使用 Yahoo Finance 核实发行方和证券；只有在公司、产品、市场或 subreddit 背景足以确认时，才将普通词语的匹配计为相关结果。
注重准确性的 Reddit 搜索：
分别使用股票代码和发行方名称查询，筛选相关 subreddit，并检查少量初始样本；优化嘈杂的搜索，而不是深入翻页查找误匹配。
可审计抽样：
记录查询设置，按帖子 ID 去重，仅获取选定主题帖的评论，并分别报告帖子和评论样本。
谨慎解读互动数据：
默认信息流可能不包含得分和评论数；metrics 模式成本更高，投票数据只是近似值，评论指标也可能只显示大型主题帖的部分内容。
匹配时间窗口的市场背景：
获取讨论期间的 Yahoo Finance 历史数据，区分历史数据与当前报价，并记录日期、间隔、币种和调整依据。
关注来源的事件背景：
有帮助时，查询股票代码相关新闻或一手申报来源，保留发布日期，不将未经核实的 Reddit 说法当作事实转述，也不暗示讨论导致价格变动。
负责任的报告：
呈现覆盖范围限制和反例，不声称抽样的 Reddit 用户代表所有投资者，也不提供买入、卖出、持有或择时建议。

name: reddit-stock-sentiment-research description: Summarize sampled Reddit discussion about a publicly traded company or ticker and, when useful, compare it with same-window Yahoo Finance price context. Use for requests like “what are Reddit investors saying about NVDA?” or “summarize bullish and bearish Reddit themes”; it is not a trading signal or price forecast.

Reddit stock sentiment research

Build a bounded, auditable brief about Reddit discussion around one publicly traded company or security. Separate what Redditors said from observed price movement, and distinguish the sampled posts from the full investor population.

Establish the security and scope

Confirm the ticker, issuer, exchange/market, and requested time window. Resolve ambiguous names or ticker strings with Yahoo Finance search and verify the selected symbol in its quote or company info.
Agree which communities matter and what question the user wants answered: discussion themes, perceived catalysts/risks, or comparison with a dated price window. If omitted, state a modest sample scope and the default market/time range.
Ticker text can be an ordinary word or a different company's symbol. Require company, product, market, or subreddit context before counting a result as relevant.

Collect public discussion

Search Reddit with the verified ticker and issuer name as separate query variants. Use a relevant subreddit filter when it improves precision; do not assume a search result page is exhaustive.
Review a small first page, record the query/filter/sort/time, and label each candidate relevant, ambiguous, or unrelated. If noise dominates, tighten the query or community instead of paging deeply through false matches.
Fetch the original post and comments only for selected relevant threads. Deduplicate by post ID. Keep post count and sampled comment count separate; repeated comments in one thread are not independent investor opinions.
Reddit's time filter is supported for top/comments sorts, not as a reliable date filter for new. For new results, inspect source dates locally and retain unknown dates as unknown. The default search feed may leave score and comment-count fields unpopulated; missing engagement is unknown, not zero.
Use Reddit's ordinary public feed by default. Engagement metrics require the more expensive metrics mode; request it only when explicitly relevant, and label the resulting counts as approximate because Reddit fuzzes voting data. Comment metric mode can expose only part of a large thread.

Add price context only when useful

Use Yahoo Finance search to resolve the security, then retrieve a quote or a historical series matching the discussion window. Record symbol, currency, interval, start/end, and whether prices are adjusted. Keep market sessions and post dates aligned; do not treat a current quote as historical context.

If news context helps explain a documented event, use the ticker news surface or a primary filing source and preserve publication dates. Do not repeat an unverified Reddit rumor as a fact or imply that discussion caused a price move.

Classify and report sampled evidence

Classify selected posts/comments by the view they actually express: positive, negative, mixed, neutral/question, or unclear. Separately tag themes such as product experience, financial performance, valuation, catalyst, risk, or speculation. Quote sparingly and link each example to its post; do not infer the author's portfolio, identity, or trading activity from a comment.

Return a brief with:

Security identity, market, discussion window, retrieval time, and Reddit queries/communities.
Relevant, ambiguous, and excluded post counts; thread and comment sample sizes.
A theme table with stance, evidence count, representative post links, and counterexamples.
Optional Yahoo Finance price context with currency, interval, exact dates, and adjustment basis.
Coverage limits and unanswered questions. Do not report an overall sentiment percentage without a clear sampled denominator.

Reddit sentiment is neither a representative poll nor a forecast. Do not recommend buying, selling, holding, or timing a security from this workflow.

Examples

scripts/crawlora.sh /yahoo-finance/search q="NVIDIA" quotes_count=5
scripts/crawlora.sh /reddit/search q="NVDA" sort=top time=month limit=10
scripts/crawlora.sh /reddit/search q="NVIDIA" subreddit=stocks sort=top time=month limit=10

# After selecting a post and confirming the symbol, substitute their identifiers:
# scripts/crawlora.sh /reddit/comments/

sort=top limit=25

# scripts/crawlora.sh /yahoo-finance/ticker//history period=1mo interval=1d

Read reference/endpoints.md for exact route parameters. Use the bundled scripts/crawlora.sh helper, which keeps CRAWLORA_API_KEY in an HTTP header.

主题需求验证
多项目体育赛事研究