メニュー
Access FinHot curated financial news feeds — RSS, JSON API, and detail pages. Use this skill to query selected/filtered financial news, hot topics, and daily digests programmatically.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | finhot |
| description | Access FinHot curated financial news feeds — RSS, JSON API, and detail pages. Use this skill to query selected/filtered financial news, hot topics, and daily digests programmatically. |
FinHot is an AI-powered financial news aggregator that processes multi-source feeds (Weibo, Xueqiu, Twitter, WeChat) through local AI enrichment (summarization, translation, scoring, embedding clustering) and exposes curated results via public endpoints.
Sources (RSSHub/Playwright/nitter)
→ Local AI pipeline (summary, translation, 6-dim scoring, embedding)
→ .finhot-cache/ (manifest.json, entries/*.json, enrichments.json)
→ Public endpoints (HTML, RSS, JSON API)
All AI processing happens locally. Public endpoints only render pre-computed results.
/public — Full reader UI with 精选/全部/热点 views/items/<id> — Standalone permalink for sharing (includes 精选理由, AI摘要, 翻译, 评分明细, 相关报道)| Endpoint | Content | Use case |
|---|---|---|
/feed.xml | 精选 (qualityScore ≥ 70) | RSS reader, daily curated feed |
/feed/all.xml | All entries | Full firehose |
/feed/daily.xml | Today's digest | Daily briefing |
RSS item links point to FinHot detail permalinks (/items/<id>). The original article URL is kept as the RSS <source> URL and on the detail page.
The public site (https://finhot.industry7view.com) is a static Cloudflare Pages deploy — the dynamic API below only runs on the local dev server. On the public site, use the pre-generated snapshot files instead (refreshed on every deploy; filter client-side with jq):
| Path | Content |
|---|---|
/api/public/items.json | Selected items snapshot (with permalink field) |
/api/public/items-all.json | Selected + watch tier snapshot |
/api/public/topics.json | Hot topic clusters |
/api/public/daily.json | Today's digest (lead + category sections) |
/api/public/daily/{date}.json | Digest archive for a specific date |
/api/public/dailies.json | Archive index (date + lead title + counts) |
/api/public/version.json | apiVersion / skillVersion / generatedAt |
/finhot-skill/SKILL.md | Agent-facing skill (canonical: skills/finhot-skill/) |
/finhot-skill/install.sh | One-line skill installer |
All API endpoints return JSON with Access-Control-Allow-Origin: *.
GET /api/public/itemsFiltered item list with enrichment data.
Parameters:
| Param | Default | Description |
|---|---|---|
filter | selected | selected / watch / all |
since | — | ISO date, return items after this time |
q | — | Full-text search (title, summary, tags) |
category | — | Filter by feed category |
limit | 50 | Max items (capped at 500) |
Response:
{
"items": [
{
"id": "...",
"title": "...",
"url": "...",
"publishedAt": "...",
"author": "...",
"feedTitle": "...",
"feedCategory": "微博",
"summary": "AI-generated content summary",
"recommendationReason": "Why this is worth reading",
"qualityScore": 82,
"selected": "selected",
"tags": ["macro", "央行"],
"translation": null,
"clusterId": "...",
"relatedEntryIds": ["...", "..."]
}
],
"total": 12,
"filter": "selected"
}
GET /api/public/item/<id>Single item full detail including content body, qualityDetails, and related entries.
Response:
{
"id": "...",
"title": "...",
"content": "<html content>",
"summary": "...",
"recommendationReason": "...",
"qualityScore": 82,
"selected": "selected",
"qualityDetails": {
"scores": {
"information_gain": 4,
"depth": 4,
"evidence": 5,
"actionability": 3,
"originality": 4,
"signal_density": 5
},
"positiveReasons": ["..."],
"negativeReasons": ["..."],
"confidence": 0.82
},
"relatedEntries": [{ "id": "...", "title": "...", "feedTitle": "...", "qualityScore": 75 }]
}
GET /api/public/topicsHot topic clusters (multi-source event aggregation). The endpoint uses the same public radar defaults as the UI: recent 3-day entries, 18-hour clustering window, and embedding cosine similarity threshold 0.78, with tag/title fallback when embeddings are unavailable.
Response:
{
"topics": [
{
"id": "cluster-xxx",
"title": "Representative headline",
"entries": ["id1", "id2", "id3"],
"entryIds": ["id1", "id2", "id3"],
"sourceCount": 3,
"sources": ["Source A", "Source B"],
"entryCount": 3,
"avgQualityScore": 78,
"qualityScore": 78,
"selected": "selected",
"earliestAt": "...",
"latestAt": "...",
"publishedAt": "...",
"summary": "...",
"tags": ["macro"]
}
]
}
GET /api/public/dailyDigest for one day: lead headline (top-scored selected item), sections grouped by feed category, plus flat selection tiers.
Parameters:
| Param | Default | Description |
|---|---|---|
date | today | ISO date (YYYY-MM-DD); also accepted as path /api/public/daily/{date} |
Response:
{
"date": "2024-01-15",
"totalEntries": 45,
"lead": { "title": "...", "qualityScore": 91, ... },
"sections": [{ "label": "微博", "items": [...] }],
"selected": [...],
"watch": [...]
}
GET /api/public/dailiesArchive index of recent daily digests: { items: [{ date, leadTitle, selectedCount }], total }. Param take (default/max 14).
GET /api/public/versionVersion info for agent self-update checks: { apiVersion, skillVersion, generatedAt, installUrl, skillUrl }. skillVersion comes from skills/finhot-skill/VERSION.
| Tier | qualityScore | Display | Meaning |
|---|---|---|---|
selected | ≥ 70 | 精选 82 | High-value, curated |
watch | 40–69 | 观察 61 | Worth monitoring |
noise | < 40 | (hidden) | Low signal |
These are intentionally separate fields. An agent should use recommendationReason to decide relevance, and summary for content understanding.
| Source | Method | Config |
|---|---|---|
RSSHub /weibo/user/{uid} | Docker localhost:1200 | |
| Xueqiu | Built-in finhot://xueqiu/{uid} scraper or RSSHub route | Playwright/Chrome and Xueqiu cookie when required |
| Twitter/X | RSSHub /twitter/user/{handle} | TWITTER_AUTH_TOKEN in Docker env |
| wechat2rss proxy | License-bound local or remote service |
.finhot-cache/
├── manifest.json # Feed registry + metadata
├── entries/
│ └── {feedId}.json # Entries per feed
└── enrichments.json # AI results keyed by entry ID
# 1. Get today's curated items
GET /api/public/items?filter=selected&since=2024-01-15
# 2. Check hot topics
GET /api/public/topics
# 3. Get detail for interesting item
GET /api/public/item/{id}
# 4. Subscribe to curated RSS
# Add /feed.xml to any RSS reader
可配置的巨潮公告分类订阅器：按分类+关键词+自选股抓取全市场披露，输出标准 Atom RSS，并把命中的官方公告作为 L3 候选(L1_L3_candidate)归档到知识库 wiki/raw/disclosures 供 review。触发：巨潮RSS、公告订阅、cninfo rss、补L3公告。
FinHot (finhot.industry7view.com) 中文金融资讯查询 Skill。当用户想知道"今天金融圈/财经有什么"、"财经日报"、"FinHot"、"金融热点"、"最近市场观点"、"雪球/微博大V最近说什么"、"财经精选"、"金融资讯"等任何中文财经资讯查询时使用。即使用户只说"今天市场有什么新闻"、"看下财经热点"，也应该触发本 Skill。Skill 直接 curl 公开静态快照 JSON 拉数据并整理成中文简报，不需要 API Key 或 MCP server。不要 undertrigger——用户问财经资讯而你不调本 Skill 就是把过时的训练数据当作今日新闻。