xauusd-calendar-descriptions

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

SKILL.md - XAUUSD Economic Calendar Descriptions (Global, Human-Readable, Research-First)

SKILL.md - XAUUSD经济日历事件描述(全球通用、易读易懂、基于调研)

Goal

目标

Generate concise, human-readable descriptions for economic calendar events in a global calendar, with optional impact notes only when relevant to XAUUSD.
为全球经济日历的事件生成简洁、易读易懂的描述,仅当事件与XAUUSD相关时才添加影响备注。

Scope

适用范围

  • Calendar includes all countries/regions.
  • Asset focus is XAUUSD.
  • Output: 1-2 sentences per event.
  • 日历覆盖所有国家/地区。
  • 聚焦资产为XAUUSD。
  • 输出要求:每个事件1-2句话。

Output Rules

输出规则

  1. At most two sentences.
  2. Sentence 1 (always): Define what the event measures/decides using precise measurement terms:
    • rate, policy decision, index, diffusion index, weekly number, % change, balance, statement, minutes, auction
  3. Sentence 2 (optional): Add only if the event can plausibly influence XAUUSD via major pricing channels.
  1. 最多两句话。
  2. 第一句(必填):使用精准的计量术语定义事件的衡量/决策内容:
    • 利率、政策决议、指数、扩散指数、周度数据、百分比变化、余额、声明、会议纪要、拍卖
  3. 第二句(可选):仅当事件可通过主要定价渠道对XAUUSD产生合理影响时添加。

Research-First Requirement (Stops Copy-Paste Notes)

基于调研的要求(避免复制粘贴备注)

Do not reuse a generic explanation across multiple events (for example, "CPI measures inflation") unless you verified it for that exact event.
Minimum bar per event:
  1. Find at least one credible source describing the indicator/decision for that exact country/issuer.
  2. Extract one short supporting excerpt (one line is enough).
  3. Only then write the 1-2 sentence note in natural English.
Source preference order:
  1. Official publisher (central bank / statistics office / ministry / exchange / treasury / agency).
  2. Official press release / PDF / page for that indicator.
  3. Reputable financial calendar / major news outlet as fallback when official sources are hard to access.
除非已针对具体事件进行验证,否则不得在多个事件中重复使用通用解释(例如“CPI衡量通胀”)。
每个事件的最低要求:
  1. 至少找到一个可信来源描述该特定国家/发行方的指标/决议。
  2. 提取一段简短的支持性摘录(一句话即可)。
  3. 之后再用自然的英文撰写1-2句话的备注。
来源优先级:
  1. 官方发布方(央行/统计局/部委/交易所/财政部/相关机构)。
  2. 该指标的官方新闻稿/PDF/专属页面。
  3. 当官方来源难以获取时,可退而求其次使用知名金融日历或主流新闻媒体。

Efficiency Without Template Notes (Cache + Official "Hub" Pages)

高效无模板备注(缓存+官方“枢纽”页面)

Goal: keep the "one-by-one" evidence standard while reducing repeated web searches and minimizing token usage.
Principles:
  1. Each event still needs its own supporting excerpt (one line) and its own final note.
  2. Efficiency comes from reusing official sources (not reusing wording).
目标:在保持“逐个验证”的证据标准同时,减少重复网页搜索并降低令牌使用量。
原则:
  1. 每个事件仍需独立的支持性摘录(一句话)和独立的最终备注。
  2. 通过复用官方来源(而非复用措辞)提升效率。

Quota-Friendly Mode (No Agents)

配额友好模式(无需多Agent)

If you cannot use multiple agents (quota constraints), do not switch to generic template notes. Keep the same per-event evidence bar and make research faster by:
  1. Using the helper script to search and cache candidate pages (
    tmp/sources/
    ).
  2. Restricting searches to official domains with
    site:
    (for example,
    site:bls.gov
    ,
    site:bea.gov
    ,
    site:census.gov
    ).
  3. Reusing cached official "hub" pages for related variants (level, MoM, YoY) while still writing each event note individually.
若因配额限制无法使用多Agent,请勿切换为通用模板备注。 保持相同的逐事件证据标准,通过以下方式加快调研速度:
  1. 使用辅助脚本搜索并缓存候选页面(存储于
    tmp/sources/
    )。
  2. 限制搜索范围为官方域名,使用
    site:
    语法(例如
    site:bls.gov
    site:bea.gov
    site:census.gov
    )。
  3. 为相关变体(同比、环比、绝对值)复用缓存的官方“枢纽”页面,同时仍需为每个事件单独撰写备注。

This Repo's Exact Workflow (What "Good" Looks Like Here)

本仓库的具体工作流程(规范示例)

When following this skill in this repository, the expected workflow is:
  1. Pick exactly one event id to work on (for example,
    USD::pce deflator::m/m
    ).
  2. Find at least one credible source line for that exact event (prefer official publisher pages).
  3. Cache sources under
    tmp/
    (gitignored), extract a one-line excerpt, and record it in
    tmp/event_notes_research_log.md
    (do not commit).
  4. Write/replace the note in
    event_notes.json
    in natural English (1-2 sentences).
  5. Run the repo quality gate (
    python scripts/calendar/build_event_notes.py
    ) and rewrite if it flags anything.
在本仓库中遵循此规范时,预期工作流程如下:
  1. 选择一个具体的事件ID进行处理(例如
    USD::pce deflator::m/m
    )。
  2. 为该具体事件找到至少一条可信来源内容(优先选择官方发布方页面)。
  3. 将来源缓存到
    tmp/
    目录(已被Git忽略),提取一句话摘录,并记录到
    tmp/event_notes_research_log.md
    (无需提交)。
  4. event_notes.json
    中用自然英文撰写/替换备注(1-2句话)。
  5. 运行仓库质量检查脚本(
    python scripts/calendar/build_event_notes.py
    ),若有报错则重写备注。

Fast, reproducible web research (recommended)

快速可复现的网页调研(推荐)

Use
r.jina.ai
for both search and fetching readable page text, then cache locally:
powershell
$ProgressPreference = 'SilentlyContinue'
New-Item -ItemType Directory -Force tmp/sources | Out-Null
使用
r.jina.ai
进行搜索并获取可读的页面文本,然后本地缓存:
powershell
$ProgressPreference = 'SilentlyContinue'
New-Item -ItemType Directory -Force tmp/sources | Out-Null

1) Search (DuckDuckGo HTML via r.jina.ai)

1) 搜索(通过r.jina.ai获取DuckDuckGo HTML结果)

$q = [uri]::EscapeDataString('BLS CPI definition prices paid by urban consumers market basket') (Invoke-WebRequest -UseBasicParsing "https://r.jina.ai/http://duckduckgo.com/html/?q=$q").Content | Set-Content -Encoding utf8 tmp/sources/ddg_cpi.md
$q = [uri]::EscapeDataString('BLS CPI definition prices paid by urban consumers market basket') (Invoke-WebRequest -UseBasicParsing "https://r.jina.ai/http://duckduckgo.com/html/?q=$q").Content | Set-Content -Encoding utf8 tmp/sources/ddg_cpi.md

2) Fetch the best official hit (again via r.jina.ai) and cache it

2) 获取最佳官方结果(再次通过r.jina.ai)并缓存

$url = 'https://www.bls.gov/cpi/' (Invoke-WebRequest -UseBasicParsing "https://r.jina.ai/$url").Content | Set-Content -Encoding utf8 tmp/sources/bls_cpi.md
$url = 'https://www.bls.gov/cpi/' (Invoke-WebRequest -UseBasicParsing "https://r.jina.ai/$url").Content | Set-Content -Encoding utf8 tmp/sources/bls_cpi.md

3) Extract one supporting excerpt line for evidence

3) 提取一条支持性摘录作为证据

Select-String -Path tmp/sources/bls_cpi.md -Pattern 'average change over time' | Select-Object -First 1

You can also use the helper script for one event id:

```powershell
python scripts/calendar/research_event_note_sources.py --event-id "USD::pce deflator::m/m"
It does not edit JSON; it only helps you find/copy evidence lines faster.
Select-String -Path tmp/sources/bls_cpi.md -Pattern 'average change over time' | Select-Object -First 1

你也可以使用辅助脚本处理单个事件ID:

```powershell
python scripts/calendar/research_event_note_sources.py --event-id "USD::pce deflator::m/m"
该脚本不会编辑JSON文件,仅帮助你更快地查找/复制证据内容。

When official sites block CLI fetching (WAF, SPA, 403)

当官方网站阻止CLI请求时(WAF、SPA、403错误)

Some official sites are hard to fetch from a CLI (CloudFront 403, "Request Rejected", single-page apps). Do not fall back to generic template notes.
Use this order:
  1. Official PDF press release / methodology note (often easier to fetch than the main site).
  2. Official API or machine-readable endpoint (JSON/CSV) if the institution provides it.
  3. An official secondary domain or a machine-readable mirror that contains the same text verbatim.
  4. Only if all official routes are blocked: a reputable secondary source (major financial calendar or major news outlet), and keep the note strictly definitional.
部分官方网站难以通过CLI请求获取内容(如CloudFront 403、“请求被拒绝”、单页应用)。请勿退而使用通用模板备注。
请按以下顺序尝试:
  1. 官方PDF新闻稿/方法论说明(通常比主站更易获取)。
  2. 若机构提供,使用官方API或机器可读端点(JSON/CSV格式)。
  3. 包含相同原文内容的官方二级域名或机器可读镜像站。
  4. 仅当所有官方渠道均被阻止时,使用可信的二级来源(知名金融日历或主流新闻媒体),且备注需严格限定为定义性内容。

Evidence logging (must do per event)

证据记录(每个事件必须执行)

Append one short block per event to
tmp/event_notes_research_log.md
:
text
EVENT: USD::cpi::m/m
SOURCE: https://www.bls.gov/cpi/
EXCERPT: "The Consumer Price Index (CPI) is a measure of the average change over time in the prices paid by urban consumers for a market basket..."
NOTE: (paste your final 1-2 sentence note)
tmp/event_notes_research_log.md
中为每个事件追加一个简短记录块:
text
EVENT: USD::cpi::m/m
SOURCE: https://www.bls.gov/cpi/
EXCERPT: "The Consumer Price Index (CPI) is a measure of the average change over time in the prices paid by urban consumers for a market basket..."
NOTE: (粘贴你最终撰写的1-2句话备注)

Repo Quality Gates (Important)

仓库质量检查(重要)

This repo enforces a couple of non-obvious checks via
python scripts/calendar/build_event_notes.py
:
  1. Banned phrases: never use
    tracks
    or
    growth indicator
    anywhere in notes.
  2. Sentence limit: a simple punctuation heuristic counts
    . ! ?
    as sentence endings.
    • Avoid abbreviations with periods like
      U.S.
      because they can be miscounted as extra sentences.
    • Prefer
      US
      instead of
      U.S.
      and keep punctuation minimal.
  3. ASCII-safe text: avoid curly quotes (use
    '
    instead of
    ) so JSON stays clean and consistent.
Recommended workflow:
  1. Identify the official publisher domain for the event (central bank, statistics office, exchange, etc.).
  2. Find 1-3 "hub" pages that cover many indicators for that publisher (glossary, methodology, release landing page, catalog page).
  3. Cache hub pages locally under
    tmp/
    (gitignored) so you can extract excerpts offline for many events.
  4. For each event, extract a one-line excerpt from cached content (or fetch a specific page if needed), then write the note.
  5. Record evidence per event (URL + excerpt) in
    tmp/event_notes_research_log.md
    (do not commit).
Optional helper script:
  • Use
    scripts/calendar/research_event_note_sources.py
    to quickly collect candidate sources and definition-like excerpt lines for one event id.
  • The script does not change any JSON; it only caches pages under
    tmp/sources/
    and prints copy-paste evidence candidates.
Examples:
powershell
python scripts/calendar/research_event_note_sources.py --event-id "USD::m2 money supply::m/m"
python scripts/calendar/research_event_note_sources.py --event-id "USD::opec monthly report::none" --max-results 8
python scripts/calendar/research_event_note_sources.py --event-id "EUR::cpi::y/y" --query "site:ec.europa.eu HICP year-on-year change definition"
PowerShell caching pattern:
powershell
$ProgressPreference = 'SilentlyContinue'
New-Item -ItemType Directory -Force tmp/sources | Out-Null
本仓库通过
python scripts/calendar/build_event_notes.py
执行以下几项非显性检查:
  1. 禁用短语:备注中不得使用
    tracks
    growth indicator
  2. 句子数量限制:通过简单的标点符号规则统计句子结尾(
    . ! ?
    )。
    • 避免使用带句号的缩写(如
      U.S.
      ),因为会被误判为额外句子。
    • 优先使用
      US
      而非
      U.S.
      ,并尽量减少标点使用。
  3. ASCII安全文本:避免使用弯引号(用
    '
    替代
    ),确保JSON文件干净一致。
推荐工作流程:
  1. 确定事件的官方发布方域名(央行、统计局、交易所等)。
  2. 找到1-3个覆盖该发布方多个指标的“枢纽”页面(术语表、方法论、发布首页、目录页)。
  3. 将枢纽页面缓存到本地
    tmp/
    目录(已被Git忽略),以便离线为多个事件提取摘录。
  4. 为每个事件从缓存内容中提取一句话摘录(或按需获取特定页面),然后撰写备注。
  5. 为每个事件记录证据(URL+摘录)到
    tmp/event_notes_research_log.md
    (无需提交)。
可选辅助脚本:
  • 使用
    scripts/calendar/research_event_note_sources.py
    快速为单个事件ID收集候选来源和类定义摘录。
  • 该脚本不会修改任何JSON文件,仅会将页面缓存到
    tmp/sources/
    目录,并打印可直接复制的候选证据。
示例:
powershell
python scripts/calendar/research_event_note_sources.py --event-id "USD::m2 money supply::m/m"
python scripts/calendar/research_event_note_sources.py --event-id "USD::opec monthly report::none" --max-results 8
python scripts/calendar/research_event_note_sources.py --event-id "EUR::cpi::y/y" --query "site:ec.europa.eu HICP year-on-year change definition"
PowerShell缓存示例:
powershell
$ProgressPreference = 'SilentlyContinue'
New-Item -ItemType Directory -Force tmp/sources | Out-Null

Cache an official hub page once

一次性缓存官方枢纽页面

$url = 'https://www.abs.gov.au/statistics/economy/price-indexes-and-inflation/consumer-price-index-australia/latest-release' $cache = 'tmp/sources/abs_cpi_latest_release.md' (Invoke-WebRequest -UseBasicParsing ("https://r.jina.ai/$url")).Content | Set-Content -Encoding utf8 $cache
$url = 'https://www.abs.gov.au/statistics/economy/price-indexes-and-inflation/consumer-price-index-australia/latest-release' $cache = 'tmp/sources/abs_cpi_latest_release.md' (Invoke-WebRequest -UseBasicParsing ("https://r.jina.ai/$url")).Content | Set-Content -Encoding utf8 $cache

For each event, extract a supporting excerpt from the cached file

为每个事件从缓存文件中提取支持性摘录

Get-Content $cache | Select-String -Pattern 'measures household inflation' -Context 0,0 | Select-Object -First 1

Search strategy (high signal, low noise):

* Prefer `site:<official-domain> <indicator name> definition` queries.
* Use `https://r.jina.ai/http://duckduckgo.com/html/?q=...` for search when direct access is blocked.
* Only fall back to non-official sources when official pages are inaccessible.

Keep diffusion-index thresholds:

* For PMI-style diffusion indexes, keep the threshold explanation when it is relevant and known (for example, `50 = expansion`).
  Do not delete this detail just to shorten the note; it materially improves usability for non-specialists.
Get-Content $cache | Select-String -Pattern 'measures household inflation' -Context 0,0 | Select-Object -First 1

搜索策略(高信号、低噪音):

* 优先使用`site:<官方域名> <指标名称> definition`格式的查询。
* 当直接访问被阻止时,使用`https://r.jina.ai/http://duckduckgo.com/html/?q=...`进行搜索。
* 仅当官方页面无法访问时,才退而使用非官方来源。

保留扩散指数阈值说明:

* 对于PMI类扩散指数,若阈值解释已知且相关,则需保留(例如`50 = 扩张`)。不要为了缩短备注而删除此细节,这对非专业用户的可用性至关重要。

Relevance Decision (When to add Sentence 2)

相关性判定(何时添加第二句)

Add an impact sentence only if at least one condition is true:
仅当至少满足以下条件之一时,才添加影响说明句:

High relevance (usually add impact)

高相关性(通常添加影响说明)

  • Major central bank policy decisions/guidance from systemically important economies (rate decision, statement, minutes, press conference, key testimony/speech)
  • Major inflation prints that can shift global rate expectations (headline/core CPI or equivalents, PCE equivalents, PPI equivalents - especially for large economies)
  • Major labor prints that move rate expectations or risk sentiment (jobs/payrolls, unemployment rate, wages; weekly claims where applicable)
  • Major activity/risk indicators from large economies (PMI/ISM equivalents, GDP, retail sales, industrial production)
  • 系统重要性经济体的主要央行政策决议/指引(利率决议、声明、会议纪要、新闻发布会、关键证词/演讲)
  • 可能改变全球利率预期的主要通胀数据( headline/core CPI或同类指标、PCE同类指标、PPI同类指标——尤其是大型经济体的数据)
  • 可能影响利率预期或风险情绪的主要就业数据(就业/非农数据、失业率、薪资;适用情况下的周度申领数据)
  • 大型经济体的主要活动/风险指标(PMI/ISM同类指标、GDP、零售销售、工业生产)

Medium relevance (add impact only if widely watched / high-impact on the calendar)

中等相关性(仅当事件被广泛关注/日历中影响较高时添加)

  • Secondary inflation/labor/activity data from mid-sized economies
  • Trade balance/current account when it is a market focus driver
  • Long-dated government bond auctions (can affect yields/rates narrative)
  • 中等经济体的次级通胀/就业/活动数据
  • 成为市场焦点驱动因素的贸易收支/经常账户数据
  • 长期政府债券拍卖(可能影响收益率/利率叙事)

Low relevance (do NOT add impact)

低相关性(不得添加影响说明)

  • Small-country second-tier releases unlikely to move global USD/yields/risk
  • Niche sector indicators with limited market attention
  • Local events with little spillover
Default behavior: If uncertain, omit Sentence 2 (definition-only is always acceptable).
  • 小国的二线数据,不太可能影响全球美元/收益率/风险情绪
  • 市场关注度有限的细分行业指标
  • 溢出效应极小的本地事件
默认规则:若不确定,请勿添加第二句(仅保留定义内容始终合规)。

Impact Sentence Rules (If Sentence 2 is used)

影响说明句规则(若使用第二句)

Impact must be conditional and relative to expectations:
  • Use: "higher-than-expected / lower-than-expected (vs forecast)"
  • For central banks: "hawkish/dovish surprise"
Impact must explain through one primary channel (pick the most relevant, do not stack multiple):
  1. USD and (real) yields / rate expectations
  2. Risk sentiment / safe-haven demand
  3. Inflation expectations feeding into yields
Keep it realistic: use can / may / often, not absolute "will".
影响说明必须是条件性的、相对于预期的:
  • 使用:“高于预期/低于预期(vs 市场预测)”
  • 对于央行事件:“鹰派/鸽派意外”
影响说明必须通过一个主要渠道解释(选择最相关的渠道,不要叠加多个):
  1. 美元与(实际)收益率/利率预期
  2. 风险情绪/避险需求
  3. 通胀预期传导至收益率
保持表述现实:使用can/may/often,而非绝对的“will”。

Banned / Anti-Robotic Rules

禁用/避免机械表述规则

  • Never use: "tracks"
  • Never call everything a: "growth indicator"
  • Never repeat the event name as its own definition
  • Never prefix events with currency codes as if currency = country
  • Avoid filler like "important indicator" or "for the economy"; say what it measures instead
  • Avoid long, encyclopedic sentences and heavy parentheses
  • 不得使用:“tracks”
  • 不得将所有指标都称为:“growth indicator”
  • 不得将事件名称作为自身定义重复使用
  • 不得将货币代码作为事件前缀(默认货币=国家)
  • 避免使用“重要指标”或“对经济而言”这类冗余表述;直接说明其衡量内容
  • 避免冗长的百科式句子和过多括号

No-Code / No-Batch-Generation Rule (Must Follow)

无代码/无批量生成规则(必须遵守)

  • Do not generate descriptions "by code." Do not output any code, pseudocode, scripts, loops, batch logic, JSON batch structures, placeholders, or variable-style tokens (for example, anything like
    {Country}
    /
    {Period}
    / bracketed template fields).
  • Each event description must read like it was individually written in natural language, even if there are thousands of events. Avoid uniform "copy-paste sentence skeletons" that clearly look programmatically stitched.
  • Do not mention the generation method (no meta text like "auto-generated", "using a template", "based on rules"). Only output the final description text.
  • 不得通过“代码生成”描述内容。不得输出任何代码、伪代码、脚本、循环、批量逻辑、JSON批量结构、占位符或变量式令牌(例如
    {Country}
    /
    {Period}
    /带括号的模板字段)。
  • 每个事件描述必须看起来像是人工单独撰写的自然语言内容,即使有成千上万个事件。避免使用明显是程序拼接的统一“复制粘贴句式框架”。
  • 不得提及生成方式(不得出现“自动生成”“使用模板”“基于规则”这类元文本)。仅输出最终的描述文本。

Low-Relevance Handling (Optional)

低相关性事件处理(可选)

If you must mention impact but relevance is low, use one restrained line only:
  • "Usually limited direct impact on XAUUSD unless it shifts broader USD, yields, or risk sentiment."
若必须提及影响但相关性较低,仅可使用一句克制的表述:
  • “通常对XAUUSD的直接影响有限,除非其显著改变整体美元、收益率或风险情绪。”

Quality Gate (Self-check every event; rewrite if any fail)

质量自检(每个事件必须执行;若违反规则则重写)

Fail if:
  • Banned phrases appear ("tracks", "growth indicator")
  • More than 2 sentences
  • Sentence 1 does not clearly define what it measures/decides
  • Sentence 2 exists but lacks higher/lower-than-expected or hawkish/dovish framing
  • Impact does not connect to USD/yields/risk/safe-haven in a plausible way
  • Text contains any placeholders/variables or looks like a batch template output
  • You did not do per-event source verification (Research-First Requirement)
若出现以下情况则判定为不合格:
  • 出现禁用短语(“tracks”“growth indicator”)
  • 句子数量超过2句
  • 第一句未清晰定义事件的衡量/决策内容
  • 存在第二句但未包含高于/低于预期或鹰派/鸽派的框架
  • 影响说明未合理关联美元/收益率/风险/避险需求
  • 文本包含任何占位符/变量或看起来像批量模板输出
  • 未执行逐事件来源验证(基于调研的要求)

Style Targets (Make it sound human)

风格目标(让表述更人性化)

  • Use simple verbs: measures, sets, reflects, shows, provides
  • Prefer concrete nouns: reading, index, policy rate, price pressures, weekly claims
  • Keep the tone trader-readable: short, direct, no fluff
  • 使用简单动词:measures、sets、reflects、shows、provides
  • 优先使用具体名词:reading、index、政策利率、价格压力、周度申领数据
  • 保持交易者友好的语气:简短、直接、无冗余内容

Examples (Global calendar, XAUUSD-focused)

示例(全球日历、聚焦XAUUSD)

  • "Eurozone CPI (YoY) measures consumer inflation. Higher-than-expected inflation can raise yield expectations and pressure XAUUSD; softer inflation can support it."
  • "Bank of Japan Rate Decision sets the policy rate. A more hawkish-than-expected outcome can lift yields and weigh on XAUUSD; a dovish surprise can support it."
  • "China PMI is a diffusion index of business activity (50 = expansion). A weaker-than-expected reading can hurt risk sentiment and support XAUUSD as a safe haven."
  • "Norway Retail Sales (MoM) measures the monthly % change in consumer spending." (definition-only)
  • “欧元区CPI(同比)衡量消费者通胀水平。高于预期的通胀可能推升收益率预期并对XAUUSD形成压力;通胀疲软则可能支撑XAUUSD。”
  • “日本央行利率决议设定政策利率。超出预期的鹰派结果可能推升收益率并打压XAUUSD;鸽派意外则可能支撑XAUUSD。”
  • “中国PMI是衡量商业活动的扩散指数(50=扩张)。低于预期的读数可能打击风险情绪,进而支撑XAUUSD的避险需求。”
  • “挪威零售销售(环比)衡量消费者支出的月度百分比变化。”(仅定义内容)

Core Execution Instruction (Use verbatim)

核心执行指令(严格遵循)

For each economic calendar event, write a short human-readable description for XAUUSD. Use at most two sentences. The first sentence must define what the event measures or decides using precise measurement terms (rate/index/diffusion index/weekly number/% change/balance/statement/minutes/auction). Before writing, do per-event web research: find at least one credible source for that exact event (prefer official publishers), extract a short supporting excerpt, and only then write the note. Do not reuse identical wording across countries unless the source supports that similarity. Use caching for efficiency: prefer official hub pages, save them under
tmp/
, and extract per-event excerpts from cached sources. This reduces search overhead without lowering the evidence standard. Only add a second sentence if the event is meaningfully relevant to XAUUSD (major central bank policy/guidance, major inflation, major labor, or major activity/risk indicators from widely watched economies, or anything that can shift global USD and yield expectations or risk sentiment). If you add impact, it must be conditional and relative to expectations: use "higher-than-expected / lower-than-expected (vs forecast)" or "hawkish/dovish surprise", and explain via USD, yields/real yields, rate expectations, or risk sentiment/safe-haven demand. Never use "tracks", never call everything a "growth indicator", never repeat the event name as its own definition, and never prefix events with currency codes as if currency = country. For PMI-style diffusion indexes, keep the threshold explanation when relevant (for example, 50 = expansion). If relevance is low, write only the definition (or optionally one very restrained line saying direct impact is usually limited). After writing, self-check and rewrite if any rule is violated. Do not output any code, placeholders, or batch-template text. Each description must read like it was individually written.
为每个经济日历事件撰写简短、易读易懂的XAUUSD相关描述。最多使用两句话。第一句必须使用精准的计量术语(rate/index/diffusion index/weekly number/% change/balance/statement/minutes/auction)定义事件的衡量或决策内容。撰写前需执行逐事件网页调研:为该具体事件找到至少一个可信来源(优先选择官方发布方),提取一段简短的支持性摘录,之后再撰写备注。除非来源支持,否则不得在不同国家的事件中复用相同措辞。为提升效率使用缓存:优先使用官方枢纽页面,将其保存到
tmp/
目录,并从缓存来源中提取逐事件摘录。这可减少搜索开销且不降低证据标准。仅当事件与XAUUSD存在实质性关联时才添加第二句(主要央行政策/指引、主要通胀、主要就业、或被广泛关注经济体的主要活动/风险指标,或任何可能改变全球美元及收益率预期或风险情绪的事件)。若添加影响说明,必须是条件性的、相对于预期的:使用“高于预期/低于预期(vs 市场预测)”或“鹰派/鸽派意外”,并通过美元、收益率/实际收益率、利率预期或风险情绪/避险需求渠道解释影响。不得使用“tracks”,不得将所有指标都称为“growth indicator”,不得将事件名称作为自身定义重复使用,不得将货币代码作为事件前缀(默认货币=国家)。对于PMI类扩散指数,若相关则保留阈值说明(例如50=扩张)。若相关性较低,仅撰写定义内容(或可选添加一句非常克制的表述说明直接影响通常有限)。撰写完成后进行自检,若违反任何规则则重写。不得输出任何代码、占位符或批量模板文本。每个描述必须看起来像是人工单独撰写的内容。