使用场景
- 伊朗： 冲突、美伊紧张局势、以色列-伊朗、中东地缘政治、油价/霍尔木兹海峡、预测市场、伊朗革命卫队、以色列国防军、停火
- Trump： Truth Social 帖子、贸易/关税政策、伊朗战争信号、能源、中国、行政命令
基础 URL
https://skill.capduck.com/{topic}/{feed}
可用主题：
iran、
trump思考方式 — 决策循环
始终从简报开始。让数据引导您的下一次调用。
不要调用所有端点。
步骤 1：评估 — 获取全局视角
curl -s "https://skill.capduck.com/iran"
简报包含局势评估（紧张程度 1-10、方向、关键变化）、观点摘要（带证据强度）、活跃话题（带信号和深入提示），以及重大事件（影响度 >= 8，过去 24 小时）。每个部分都告诉您接下来应该关注什么。
步骤 2：调查 — 循着线索
- 话题提到关联事件 -> 使用
/events?category=CONFLICT 查看证据链
- 某主张证据薄弱 -> 使用
/posts?source_type=osint 查看 OSINT 是否确认
- 主流叙事不清晰 -> 使用
/news?region=us 查看美国媒体如何报道
- 话题关联预测市场 -> 使用
/polymarket 查看概率和趋势
- Trump 发布关于伊朗的内容 -> 使用
/trump?tags=iran-war 查看他的原话
步骤 3：交叉引用 — 三角验证
/posts?source_type=state_media vs
osint vs
/news
- 使用
/notable 了解某来源属于哪一方
- 使用
/entities?q=name 检查不熟悉来源的可信度
- 检查事件中的
source_count — 1 个来源 = 谣言，5+ 个来源 = 已确认
- 使用
/signals?ids=id1,id2 批量查询话题信号的原始数据
步骤 4：预测 — 接下来会发生什么
- 查看活跃话题中的 关注项：可能性方向（上升/稳定/下降）、下一个里程碑、决策者
- 查看
/polymarket 获取市场隐含概率和趋势方向
- 当 polymarket 与简报评估不一致时，其中一个错了 — 这是调查价值最高的地方
伊朗端点
以下所有路径相对于 https://skill.capduck.com。
/{topic} — 简报（默认）
AI 生成的情况简报，每小时更新。包含：
- 局势评估 — 紧张程度（1-10）、方向、24 小时对比、关键变化
- 观点摘要 — 4 个视角，附带证据强度（强/中/弱）
- 活跃话题 — 3-5 个问题话题，带关联信号、关注项、polymarket 引用
- 重大事件 — 过去 24 小时内影响度 >= 8 的事件，带来源归属
/{topic}/events — 事件时间线
带影响评分和多来源归属的结构化事件。
| 参数 | 描述 | 默认值 |
|---|
impact | 最低影响评分 (1-10) | 0 |
category | CONFLICT, DIPLOMACY, POLITICS, ECONOMY, PROTESTS | 全部 |
hours | 滚动时间窗口（小时） | 48 |
since | ISO 8601 时间戳（覆盖 hours） | — |
limit | 最大结果数 | 50 |
每个事件包含：
id、
impact、
confidence、
sentiment、
category、
title /
title_zh、
summary /
summary_zh、
source_details[]（含
display_name、
entity_slug、
platform、
url）、
source_count、
published_at。
/{topic}/posts — 社交动态
统一的 Twitter + Telegram 帖子，带实体丰富数据。
| 参数 | 描述 | 默认值 |
|---|
source_type | news_agency, state_media, osint, gov_military, journalist, think_tank, other | 全部 |
category | 作者类别文本筛选 | 全部 |
platform | twitter, telegram | 全部 |
event | 事件 UUID -> 通过来源 URL 关联帖子 | — |
since | ISO 8601 时间戳 | 48 小时前 |
limit | 最大结果数 | 30 |
/{topic}/news — 主流媒体
18+ 家主要媒体的报道（路透社、美联社、CNN、纽约时报、彭博社、卫报、BBC、半岛电视台等）。
| 参数 | 描述 | 默认值 |
|---|
region | intl, us, mideast, israel, persian, asia, russia, china, other | 全部 |
language | en, fa | 全部 |
event | 事件 UUID -> 关联文章 | — |
since | ISO 8601 时间戳 | 48 小时前 |
limit | 最大结果数 | 30 |
/{topic}/polymarket — 预测市场
伊朗相关预测市场的当前概率、趋势方向和历史价格区间。每个条件包含：
label/
label_zh、
yes_price（0-1）、带百分比变化的趋势箭头、历史最小-最大区间。
/{topic}/notable — 精选关键实体
约 90 个精选关键来源，按角色分组：
| 分组 | 代表内容 |
|---|
| 伊朗官方 | Khamenei.ir、IRNA、Fars、Tasnim — 政权的声音 |
| 以色列及军事 | 以色列国防军、内塔尼亚胡、耶路撒冷邮报 — 以色列视角 |
| 美国政府 | 国务院、中央司令部、白宫、Trump — 美国政策信号 |
| 国际组织 | 国际原子能机构、北约、联合国 — 多边立场 |
| 关键媒体 — 通讯社及全球 | 路透社、美联社、彭博社、纽约时报、卫报 — 基准叙事 |
| 关键媒体 — 地区及专业 | Iran International、Al-Monitor、Bellingcat — 专业深度 |
| OSINT 及预测 | Sentdefender、OSINT613、Polymarket — 实时地面真相 |
| 智库 | 大西洋理事会、布鲁金斯学会、外交关系委员会 — 分析框架 |
| 关键活动人士及反对派 | 马西赫·阿林贾德、礼萨·帕尔维、伊朗全国抵抗委员会 — 海外/反对派声音 |
/{topic}/entities — 完整来源目录
800+ 追踪实体。也可通过别名
/sources 访问。
| 参数 | 描述 | 默认值 |
|---|
category | 按实体类别筛选 | 全部 |
q | 文本搜索（名称 + 描述） | — |
limit | 最大结果数 | 100 |
Trump 端点
/trump — 完整 Trump Truth Social 动态，按主题自动标记。
| 参数 | 描述 | 默认值 |
|---|
tags | 逗号分隔的主题筛选（iran-war、energy、trade、china、executive-order 等） | 全部 |
limit | 最大结果数 | 50（最大：200） |
since | ISO 8601 或 Unix 时间戳 | — |
cursor | 分页游标（来自 meta.next_cursor） | — |
lang | zh 或 en | zh |
offset | 分页偏移 | 0 |
每个帖子包含：
id、
author_name、
author_handle、
content（英文）、
content_zh、
tags[]、
source_url、
platform（truth-social）、
media_url、
published_at。
关键字典
事件类别： CONFLICT · DIPLOMACY · POLITICS · ECONOMY · PROTESTS
事件情绪： NEGATIVE · POSITIVE · NEUTRAL · MIXED
source_type（帖子）：news_agency · state_media · osint · gov_military · journalist · think_tank · other
新闻区域： intl · us · mideast · israel · persian · asia · russia · china · other
帖子平台： twitter · telegram · truth-social（仅 Trump）
Trump 标签： iran-war · energy · trade · china · executive-order ·（自动生成，不完全列举）
实体关系
简报 -> thread.signals[].id -> 事件 ID -> /posts?event={id} 或 /news?event={id}
事件 -> source_details[].entity_slug -> /entities 或 /notable
帖子 -> source_type（从 author_category 映射）
话题 -> polymarket_ref -> /polymarket
Trump -> tags[] -> 与 /events 或 /posts 交叉引用
关键调查路径：
- 简报 -> 话题 -> 事件 -> 来源 -> 实体："此话题声称 X" -> 检查关联事件 -> 有多少来源？-> 他们是谁？-> 属于哪一方？
- 事件 -> 帖子/新闻：使用
/posts?event={id} 或 /news?event={id} 查找与特定事件关联的原始内容
- 话题 -> Polymarket：话题有 polymarket_ref -> 检查
/polymarket -> 市场概率是否与 AI 评估一致？
- 弱证据 -> 交叉引用：主张只有 1 个来源或来自娱乐/社交账户 -> 检查
/news 的主流报道 -> 检查 /posts?source_type=osint -> 如果两者都未确认，标记为未验证
- Trump 信号 -> 伊朗影响：
/trump?tags=iran-war -> 检查事件是否证实 -> /events?category=DIPLOMACY -> 评估实际影响与言论是否一致
使用场景
"伊朗最新情况是什么？" — 调用 /iran -> 总结局势评估 + 前 2-3 个话题。保持在 300 字以内。以紧张程度和方向开头。
"Trump 刚刚对伊朗说了什么？" — 调用 /trump?tags=iran-war&limit=5 -> 然后交叉检查 /events?category=DIPLOMACY&hours=6 看言论是否与实际发展一致。
"这个来源可信吗？" — 首先调用 /notable（快速、精选）。如果找不到，回退到 /entities?q=name。报告他们属于哪一方以及其类别。
"市场怎么说？" — 调用 /polymarket -> 将市场隐含概率与简报的话题评估进行比较。明确标记分歧。
"深入了解特定事件" — 从简报或 /events 获取事件 ID -> 调用 /posts?event={id} 和 /news?event={id} -> 三角验证来源数量和来源类型。
输出指南
- 先给出判断，再提供证据。 不要倾倒原始数据 — 先综合，后引用。
- 标记置信度。 使用"已确认（N 个来源）" / "已报道但未验证（单一来源）" / "谣言"，基于
source_count。
- 标记弱证据。 如果某主张只有 1 个来源或来自娱乐/社交账户，明确说明。
- 保持简洁。 默认为 200-400 字，除非用户要求详细。
- 双语更有帮助时。 关键术语和引用可包含英文和中文（当数据提供翻译时）。
- 永远不要将 AI 分析作为事实呈现。 简报是 AI 生成的 — 用"简报评估..."而非"情况是..."来表述。
关键原则
简报是地图。其他端点是领土。
不要将所有数据倾倒给用户 — 使用简报识别重要内容，只调查相关内容，然后综合判断。
备注
- 数据来自 iranmonitor.org（800+ 来源）、Polymarket CLOB API、Truth Social
- AI 分析每小时更新，数据动态每 5 分钟更新
- 所有响应均为 Markdown，数据提供中文翻译时包含中文翻译
- 48 小时滚动数据窗口，按事件/帖子 ID 去重
- Trump 动态：按主题自动标记，双语（英文 + 中文）
When to use
- Iran: conflict, US-Iran tensions, Israel-Iran, Middle East geopolitics, oil/Hormuz, prediction markets, IRGC, IDF, ceasefire
- Trump: Truth Social posts, trade/tariff policy, Iran-war signals, energy, China, executive orders
Base URL
https://skill.capduck.com/{topic}/{feed}
Available topics: iran, trump
How to think — the decision loop
Always start from the briefing. Let the data guide your next call. Do NOT call all endpoints.
Step 1: Assess — get the big picture
curl -s "https://skill.capduck.com/iran"
The briefing contains situation assessment (tension 1-10, direction, key deltas), perspective summaries (with evidence strength), active threads (with signals and drill-down hints), and top events (impact >= 8, last 24h). Each section tells you where to look next.
Step 2: Investigate — follow the scent
- A thread mentions linked events ->
/events?category=CONFLICT to read the evidence chain
- A claim has weak evidence ->
/posts?source_type=osint to see if OSINT confirms
- Mainstream narrative unclear ->
/news?region=us to see how US outlets frame it
- A thread links to prediction markets ->
/polymarket for probability and trend
- Trump posted about Iran ->
/trump?tags=iran-war to see his exact words
Step 3: Cross-reference — triangulate
- Compare official sources vs OSINT vs mainstream:
/posts?source_type=state_media vs osint vs /news
- Use
/notable to understand who a source is and which camp they belong to
- Use
/entities?q=name to check credibility of an unfamiliar source
- Check
source_count in events — 1 source = rumor, 5+ sources = confirmed
- Use
/signals?ids=id1,id2 to batch-lookup raw data for thread signals
Step 4: Project — what happens next
- Look at Watch items in Active Threads: likelihood direction (rising/stable/falling), next milestones, decision makers
- Check
/polymarket for market-implied probabilities and trend direction
- When polymarket diverges from briefing assessment, one of them is wrong — that's where investigation value is highest
Iran Endpoints
All paths below are relative to https://skill.capduck.com.
/{topic} — Briefing (default)
AI-generated situation briefing, updated hourly. Contains:
- Situation Assessment — tension level (1-10), direction, 24h comparison, key deltas
- Perspective Summaries — 4 viewpoints with evidence strength (strong/moderate/weak)
- Active Threads — 3-5 issue threads with linked signals, watch items, polymarket refs
- Top Events — impact >= 8 from last 24h, with source attribution
/{topic}/events — Event timeline
Structured events with impact scoring and multi-source attribution.
| Param | Description | Default |
|---|
impact | Minimum impact score (1-10) | 0 |
category | CONFLICT, DIPLOMACY, POLITICS, ECONOMY, PROTESTS | all |
hours | Rolling window in hours | 48 |
since | ISO 8601 timestamp (overrides hours) | — |
limit | Max results | 50 |
Each event includes:
id,
impact,
confidence,
sentiment,
category,
title /
title_zh,
summary /
summary_zh,
source_details[] (with
display_name,
entity_slug,
platform,
url),
source_count,
published_at.
/{topic}/posts — Social feed
Unified Twitter + Telegram posts with entity enrichment.
| Param | Description | Default |
|---|
source_type | news_agency, state_media, osint, gov_military, journalist, think_tank, other | all |
category | Author category text filter | all |
platform | twitter, telegram | all |
event | Event UUID -> linked posts via source URLs | — |
since | ISO 8601 timestamp | 48h ago |
limit | Max results | 30 |
/{topic}/news — Mainstream media
Coverage from 18+ major outlets (Reuters, AP, CNN, NYT, Bloomberg, Guardian, BBC, Al Jazeera, etc.).
| Param | Description | Default |
|---|
region | intl, us, mideast, israel, persian, asia, russia, china, other | all |
language | en, fa | all |
event | Event UUID -> linked articles | — |
since | ISO 8601 timestamp | 48h ago |
limit | Max results | 30 |
/{topic}/polymarket — Prediction markets
Current probabilities, trend direction, and historical price ranges for Iran-related prediction markets.
Each condition includes: label/label_zh, yes_price (0-1), trend arrow with % change, historical min-max range.
/{topic}/notable — Curated key entities
~90 curated key sources grouped by role:
| Group | What it represents |
|---|
| Iran Official | Khamenei.ir, IRNA, Fars, Tasnim — the regime's voice |
| Israel & Military | IDF, Netanyahu, Haaretz, Jerusalem Post — Israeli perspective |
| US Government | State Dept, CENTCOM, White House, Trump — US policy signals |
| International Organizations | IAEA, NATO, UN — multilateral stance |
| Key Media — Wire & Global | Reuters, AP, Bloomberg, NYT, Guardian — baseline narrative |
| Key Media — Regional & Specialist | Iran International, Al-Monitor, Bellingcat — specialist depth |
| OSINT & Prediction | Sentdefender, OSINT613, Polymarket — real-time ground truth |
| Think Tanks | Atlantic Council, Brookings, CFR — analytical framing |
| Key Activists & Opposition | Masih Alinejad, Reza Pahlavi, NCRI — diaspora/opposition voice |
/{topic}/entities — Full source directory
800+ tracked entities. Also accessible via alias /sources.
| Param | Description | Default |
|---|
category | Filter by entity category | all |
q | Text search (name + description) | — |
limit | Max results | 100 |
Trump Endpoint
/trump — Full Trump Truth Social feed, auto-tagged by topic.
| Param | Description | Default |
|---|
tags | Comma-separated topic filter (iran-war, energy, trade, china, executive-order, etc.) | all |
limit | Max results | 50 (max: 200) |
since | ISO 8601 or Unix timestamp | — |
cursor | Pagination cursor (from meta.next_cursor) | — |
lang | zh or en | zh |
offset | Pagination offset | 0 |
Each post includes:
id,
author_name,
author_handle,
content (EN),
content_zh,
tags[],
source_url,
platform (truth-social),
media_url,
published_at.
Key dictionaries
Event category: CONFLICT · DIPLOMACY · POLITICS · ECONOMY · PROTESTS
Event sentiment: NEGATIVE · POSITIVE · NEUTRAL · MIXED
source_type (on posts): news_agency · state_media · osint · gov_military · journalist · think_tank · other
News region: intl · us · mideast · israel · persian · asia · russia · china · other
Post platform: twitter · telegram · truth-social (Trump only)
Trump tags: iran-war · energy · trade · china · executive-order · (auto-generated, non-exhaustive)
Entity relationships
Briefing -> thread.signals[].id -> Event ID -> /posts?event={id} or /news?event={id}
Event -> source_details[].entity_slug -> /entities or /notable
Post -> source_type (mapped from author_category)
Thread -> polymarket_ref -> /polymarket
Trump -> tags[] -> cross-reference with /events or /posts
Key investigation paths:
- Briefing -> Thread -> Event -> Sources -> Entity: "This thread claims X" -> check linked events -> how many sources? -> who are they? -> what camp?
- Event -> Posts/News: use
/posts?event={id} or /news?event={id} to find original content linked to a specific event
- Thread -> Polymarket: thread has polymarket_ref -> check
/polymarket -> does market probability align with AI assessment?
- Weak evidence -> Cross-reference: claim has weak evidence -> check
/news for mainstream coverage -> check /posts?source_type=osint -> if neither confirms, flag as unverified
- Trump signal -> Iran impact:
/trump?tags=iran-war -> check if events corroborate -> /events?category=DIPLOMACY -> assess real impact vs rhetoric
Usage scenarios
"What's the latest on Iran?" — Call /iran -> summarize situation assessment + top 2-3 threads. Keep it under 300 words. Lead with tension level and direction.
"What did Trump just say about Iran?" — Call /trump?tags=iran-war&limit=5 -> then cross-check /events?category=DIPLOMACY&hours=6 to see if rhetoric matches real developments.
"Is this source credible?" — Call /notable first (fast, curated). If not found, fall back to /entities?q=name. Report which camp they belong to and their category.
"What are the markets saying?" — Call /polymarket -> compare market-implied probability with briefing's thread assessment. Flag divergences explicitly.
"Deep dive on a specific event" — Get event ID from briefing or /events -> call /posts?event={id} and /news?event={id} -> triangulate source count and source types.
Output guidelines
- Lead with judgment, then evidence. Don't dump raw data — synthesize first, cite after.
- Mark confidence. Use "confirmed (N sources)" / "reported but unverified (single source)" / "rumor" based on
source_count.
- Flag weak evidence. If a claim has only 1 source or comes from entertainment/social accounts, say so explicitly.
- Keep it concise. Default to 200-400 words unless the user asks for detail.
- Bilingual when helpful. Key terms and quotes can include both English and Chinese where the data provides translations.
- Never present AI analysis as fact. The briefing is AI-generated — frame it as "the briefing assesses..." not "the situation is...".
Key principle
The briefing is the map. The other endpoints are the territory. Don't dump all data on the user — use the briefing to identify what matters, investigate only what's relevant, and synthesize a judgment.
Notes
- Data from iranmonitor.org (800+ sources), Polymarket CLOB API, Truth Social
- AI analysis hourly, data feeds every 5 minutes
- All responses are Markdown with Chinese translations where available
- 48-hour rolling data window with deduplication by event/post ID
- Trump feed: auto-tagged by topic, bilingual (EN + ZH)
by