yahoo.co.jp

homepage-news-trends-weather

Installation

Adds this website's skill for your agents

 

Summary

Briefing of the four above-the-fold modules on the Yahoo! JAPAN homepage: 8 main news topics, 5 realtime trending search keywords with stay/rise/fall direction, today/tomorrow weather for the IP-defaulted ward, and the day's NPB scoreboard. Single static HTML fetch — no JS, no auth, no anti-bot.

FIG. 01
FIG. 02
FIG. 03
FIG. 04
FIG. 05
SKILL.md
239 lines

Yahoo! JAPAN Homepage Briefing

Purpose

Return a compact briefing of the four "above-the-fold" modules on the Yahoo! JAPAN homepage (https://www.yahoo.co.jp/): the 8 main news topics (主要 ニュース), the 5 realtime trending search keywords (リアルタイム検索で話題のキーワード) with their trend direction, the today/tomorrow weather summary for the IP-defaulted location, and the sports scoreboard (NPB baseball + J.League soccer matches scheduled for the day). Read-only — never logs in, follows pickup/article links, or alters any state.

When to Use

  • Daily "Japanese news at a glance" briefings for a user-facing agent.
  • Detecting breaking-news pickup items by polling the top 8 (each pickup/N ID is unique and monotonically increases).
  • Surfacing realtime trending search terms (the closest Japanese-language analogue to "what is X-trending in Japan right now") with their stay/rise/fall direction.
  • Pulling NPB/J-League scheduled or in-progress games for the day without hitting the deeper Sportsnavi sub-sites.
  • Any workflow that previously scraped multiple Yahoo! JAPAN subdomain pages (news.yahoo.co.jp + weather.yahoo.co.jp + search.yahoo.co.jp/realtime + sports.yahoo.co.jp) when a single homepage request returns all four blocks already aggregated.

Workflow

The Yahoo! JAPAN homepage is fully server-side rendered — every value visible above the fold is already in the initial HTML response. There is no GraphQL endpoint, no __NEXT_DATA__ blob, no client-side hydration. A single unauthenticated GET https://www.yahoo.co.jp/ returns all four modules. No anti-bot challenge, no JavaScript execution required, no captcha, no Akamai. Stealth and residential proxies are not required. Both bare HTTP and residential-proxy fetches return identical structural data (only the IP-defaulted weather ward differs — see Gotchas).

Recommended path — HTTP fetch + HTML parse

  1. Fetch the homepage with any standard HTTP client (the response is brotli-compressed; modern clients handle this automatically). No headers beyond a normal User-Agent are needed.

    browse cloud fetch https://www.yahoo.co.jp/
# or any equivalent: curl, node fetch, python requests, etc.

    The response is ~150 KB of HTML, status 200, Content-Type: text/html; charset=UTF-8. Two Set-Cookie headers (A=…, B=…) are returned but do not need to be persisted for subsequent requests.

  2. Locate the 4 module containers. Each module is anchored by a Japanese <h1> whose text content (whitespace-stripped) matches one of these strings. The CSS class names on wrapper <div>/<section> elements are CSS-modules hashes (e.g. _2pjWfyGnbTPxsLzERUiAmE) that rotate on every release — do not match by class name; match by heading text, then walk up to the nearest <section> or <article> ancestor.

    • 主要ニュース → 8 news topics (wraps <section>)
    • リアルタイム検索で話題のキーワード → 5 trending keywords (wraps <section>)
    • 今日明日の天気 → weather block (wraps <article>)
    • スコアボード → sports scoreboard (wraps <article>)

  3. Parse the 主要 ニュース block. Inside the section, every a[href*="news.yahoo.co.jp/pickup/"] is a topic link (exactly 8 of them in this section — there are also 64+ pickup URLs scattered elsewhere on the page in "おすすめの記事" and "もっと見る" lists, so always scope the selector to the section ancestor, never the whole document).

    • Headline + comment count + NEW marker are concatenated in the anchor's text. Strip the trailing comment count (regex (\d+)$) and the literal "NEW" tag (regex NEW(\d+)$) to recover the clean headline. The "NEW" suffix indicates the topic was posted in the last few hours.
    • Pickup ID: parse the trailing integer from the href path. URLs are stable canonical IDs (https://news.yahoo.co.jp/pickup/6580871).
    • Update timestamp: the section's text contains a string matching \d+/\d+\([日月火水木金土]\)\s*\d+:\d+更新 (e.g. 5/20(水) 6:23更新). This is JST and updates every ~5 minutes.

  4. Parse the リアルタイム検索 block. Inside the section, every a[href*="search.yahoo.co.jp/realtime/search?rkf=1"] is a trending keyword link (5 of them). The link text is a concatenation of (in order): rank digit, trend-direction marker (stay / rise / fall), keyword, optional 写真あり photo flag, then optional 関連ワード{related keywords concatenated with no delimiter}.

    • Keyword (clean): URL-decode the p= query-string parameter of the anchor's href — this gives the keyword unambiguously, free of the concatenated trend tags and rank prefix.
    • Rank: the first <span> child of the anchor contains the rank as plain text ("1" … "5").
    • Trend direction: the third <span> child contains literally stay, rise, or fall.
    • Has photo: presence of the text 写真あり in the link text indicates an attached trending image.
    • Related words: extract by string-removing the rank + trend + keyword + 写真あり from the anchor text; whatever remains after 関連ワード is the related-words blob (no delimiter — Japanese search terms are usually short, so a heuristic split on katakana/hiragana/kanji boundaries works, or just emit the raw blob).
    • Update timestamp: the section's text contains a string matching \d+:\d+更新 (e.g. 6:35更新). HH:MM only (date is implicit = today JST).

  5. Parse the 今日明日の天気 block. The article contains:

    • Location name: an a[href*="weather.yahoo.co.jp/weather/jp/"] link whose text is only the ward name (e.g. 港区 / 新宿区) — distinguishable from the weather-data links by having no ℃ or % in its text. The href encodes the location IDs: /jp/{prefectureCode}/{areaCode}/{wardCode}.html.
    • Today / tomorrow forecasts: two more weather links whose text matches the pattern (今日|明日)の天気最高気温\d+℃最低気温\d+℃降水確率\d+%. Tomorrow's link has the fragment #yjw_pinpoint_tomorrow.
    • Weather icon / condition string: each forecast link contains a child <img> whose alt attribute is the Japanese weather phrase (晴, 雨, 晴のち雨, くもり時々雨, etc.) and whose src ends in /general/next/{code}_day.png — the {code} is a stable Yahoo internal weather-state ID (e.g. 114_day = 晴のち雨, 300_day = 雨). Extract the alt for human-readable condition.
    • Pollen forecast (花粉予報): an adjacent a[href*="weather/pollen/"] link with text like 花粉予報少ない (少ない/やや多い/多い/非常に多い). Present year-round but the level is most meaningful Feb–May.

  6. Parse the スコアボード block. Inside the article:

    • Date label: text matches \d+/\d+（[日月火水木金土]）の試合 — note the full-width parentheses （）. May be absent on off-days (no games scheduled).
    • Sport tabs: the heading row contains anchors a[href="https://baseball.yahoo.co.jp/npb/"] (プロ野球) and a[href*="soccer.yahoo.co.jp/jleague"] (Jリーグ). The active tab's content is what's rendered — by default this is プロ野球 (NPB baseball). To get J.League fixtures you must hit a sub-page on soccer.yahoo.co.jp/jleague — the homepage embed always shows the default tab only.
    • Per-game block: each game is a <dl> containing two a[href*="baseball.yahoo.co.jp/npb/teams/"] anchors (home team, then away team — order corresponds to "home 対 away") separated by a <span> with text 対, followed by an a[href*="baseball.yahoo.co.jp/npb/game/"] anchor whose text is the start time (HH:MM JST) and a second 見どころ (preview) anchor pointing to the same game URL. There are typically 5 games per day (6 NPB teams × 2 leagues split into 3 + 3 pairings; on inter-league weeks or off-days the count varies).
    • Game ID: the trailing path segment of the game URL (e.g. 2021038884) is a stable Yahoo NPB game ID.
    • Live state: a game's anchor text changes from a start-time string (e.g. 18:00) to a score string (e.g. 3 - 2) and finally to 試合終了 once it concludes. Detect the format with a regex check — ^\d{1,2}:\d{2}$ = scheduled, ^\d+\s*-\s*\d+$ = in-progress or final, presence of 試合終了 text in the dl = final.

  7. (Optional) For deeper details — pickup story body, full weather forecast for a custom ward, J.League fixtures — follow the link URLs surfaced above. These are out of scope for the homepage-briefing skill but the canonical URLs are useful for chaining.

Browser fallback

If for any reason the static HTML fetch is unavailable (network restriction, IP-blocked, etc.), the same selectors work in a browser session. No stealth or residential proxy is needed.

sid=$(browse cloud sessions create --keep-alive | node -e "let s='';process.stdin.on('data',c=>s+=c).on('end',()=>process.stdout.write(JSON.parse(s).id))")
export BROWSE_SESSION="$sid"
browse open --remote https://www.yahoo.co.jp/
browse wait --remote timeout 1500    # generous; the page is server-rendered, no hydration to wait for
browse get --remote markdown body    # captures all four sections in one snapshot
# Or use `browse eval` with the heading-text-anchored selectors above for structured extraction.
browse cloud sessions update "$sid" --status REQUEST_RELEASE

The browser path is reliable but ~10× slower and costs Browserbase session-time. Prefer the static-fetch path unless you need to interact with the page.

Site-Specific Gotchas

  • Weather location is determined by the request's source IP geolocation, not by any URL param or cookie on the homepage. A US-datacenter IP defaults to 13/4410/13103 (港区, Minato Ward, Tokyo); a Japanese residential proxy IP defaults to 13/4410/13104 (新宿区, Shinjuku Ward, Tokyo) — both inside central Tokyo. Anonymous unauthenticated users cannot pick a custom city from the homepage block without logging in. If a specific city is needed, fetch https://weather.yahoo.co.jp/weather/jp/{pref}/{area}/{ward}.html directly with the desired location codes — that's a different skill, not solvable from the homepage. Always emit the location name (e.g. 港区) alongside the temperatures so downstream consumers know what locality the numbers refer to.
  • CSS class names are hashed and rotate on every release (e.g. _2pjWfyGnbTPxsLzERUiAmE, JP_5HSOWr6XeS0joNfYde). Never select by class. Always anchor on Japanese heading text (主要ニュース, リアルタイム検索で話題のキーワード, 今日明日の天気, スコアボード) then walk to the nearest <section>/<article>. The heading strings themselves are stable.
  • Pickup URLs appear ~64 times across the homepage HTML, but only 8 are in the 主要 ニュース section. The rest are "おすすめの記事" (recommended/sponsored), "もっと見る" preview lists, and category headers. Scope the selector to the 主要ニュース section ancestor; do not query the whole document.
  • Topic link text concatenates headline + optional NEW + comment count with no delimiter (e.g. 行方不明の小4女児 23年続く捜査NEW28). Comment count is the trailing integer; NEW indicates a topic posted in the last few hours. Strip both to recover the clean headline.
  • Realtime keyword anchors include a "pickup" feature article as the first realtime-search link in the HTML (e.g. https://search.yahoo.co.jp/realtime/search/pickup/70093) — this is not one of the 5 trending keywords; it's a sponsored editorial card. Filter on href*="search.yahoo.co.jp/realtime/search?rkf=1" (note ?rkf=1) to get only the 5 ranked trending entries, not the pickup feature.
  • The trend-direction marker (stay/rise/fall) is the literal English text inside the third <span> of each keyword anchor — it's rendered as a colored arrow icon visually but the underlying text is the English word. There is no Japanese equivalent in the DOM.
  • The p= URL-encoded query parameter is the canonical way to recover the keyword. The anchor's visible text is a concatenation of rank + trend + keyword + photo flag + related-words with no separator; URL-decoding p= gives the keyword unambiguously.
  • Scoreboard default tab is NPB baseball. The J.League tab (Jリーグ) is rendered in the HTML as a navigation link, but the embedded scoreboard content always shows プロ野球 unless a user clicks the tab (and that state is not URL-persistent — opening the homepage fresh always lands on プロ野球). If you need J.League fixtures, navigate to https://soccer.yahoo.co.jp/jleague directly.
  • Scoreboard date label uses full-width parentheses (5/20（水）の試合 with （）, not ()). Match accordingly.
  • No games scheduled = no scoreboard block. On scheduled NPB off-days (Mondays during regular season, all-star break, season boundary days), the scoreboard block may be absent or show "本日試合なし" (no games today). Don't error; emit games: [].
  • All times are JST (UTC+9). The homepage does not localize for the viewer's timezone. Update timestamps (6:23更新, 6:35更新) are JST clock times.
  • No __NEXT_DATA__, no GraphQL, no internal API. The homepage is plain server-rendered HTML — there is no faster machine-readable endpoint to chase. The HTML parse is the optimal path.
  • No rate limit observed at 1 req/min. Yahoo! JAPAN sets only a session cookie (A=…) which is not required for subsequent requests. Avoid hammering — the homepage refreshes its data on the server every ~5 minutes, so polling faster than that yields no new data.
  • Browser fetch from a US-region Browserbase session works fine (no Akamai, no JS-challenge, no captcha) — but the served weather location reflects the session's exit IP. If you need to match a Japanese user's experience exactly (e.g. for QA), use --proxies to get a Japanese residential IP; otherwise omit and save the cost.

Expected Output

{
  "fetched_at_utc": "2026-05-19T21:46:00Z",
  "homepage_url": "https://www.yahoo.co.jp/",
  "news": {
    "updated_jst": "5/20(水) 6:23",
    "topics": [
      {
        "rank": 1,
        "headline": "AIミュトスに危機感 政府が対応案",
        "is_new": false,
        "comment_count": 596,
        "pickup_id": "6580871",
        "url": "https://news.yahoo.co.jp/pickup/6580871"
      },
      {
        "rank": 3,
        "headline": "行方不明の小4女児 23年続く捜査",
        "is_new": true,
        "comment_count": 28,
        "pickup_id": "6580884",
        "url": "https://news.yahoo.co.jp/pickup/6580884"
      }
    ]
  },
  "trending_keywords": {
    "updated_jst": "6:35",
    "keywords": [
      {
        "rank": 1,
        "keyword": "アーセナル優勝",
        "trend": "stay",
        "has_photo": true,
        "related_words_raw": "本当に優勝22年ぶりArsenal",
        "search_url": "https://search.yahoo.co.jp/realtime/search?rkf=1&p=%E3%82%A2%E3%83%BC%E3%82%BB%E3%83%8A%E3%83%AB%E5%84%AA%E5%8B%9D"
      },
      {
        "rank": 4,
        "keyword": "学マ水曜日",
        "trend": "rise",
        "has_photo": false,
        "related_words_raw": null,
        "search_url": "https://search.yahoo.co.jp/realtime/search?rkf=1&p=%E5%AD%A6%E3%83%9E%E6%B0%B4%E6%9B%9C%E6%97%A5"
      }
    ]
  },
  "weather": {
    "location_name": "港区",
    "location_path": "/13/4410/13103",
    "today": {
      "condition": "晴のち雨",
      "icon_code": "114_day",
      "high_c": 30,
      "low_c": 20,
      "precip_pct": 50
    },
    "tomorrow": {
      "condition": "雨",
      "icon_code": "300_day",
      "high_c": 22,
      "low_c": 16,
      "precip_pct": 80
    },
    "pollen": "少ない"
  },
  "scoreboard": {
    "sport": "npb_baseball",
    "date_label": "5/20（水）の試合",
    "games": [
      {
        "game_id": "2021038884",
        "home_team": "阪神",
        "away_team": "中日",
        "start_time_jst": "18:00",
        "status": "scheduled",
        "score": null,
        "url": "https://baseball.yahoo.co.jp/npb/game/2021038884/index"
      },
      {
        "game_id": "2021038888",
        "home_team": "オリックス",
        "away_team": "ソフトバンク",
        "start_time_jst": "18:00",
        "status": "scheduled",
        "score": null,
        "url": "https://baseball.yahoo.co.jp/npb/game/2021038888/index"
      }
    ]
  }
}

Alternate scoreboard shapes:

// Game in progress
{ "game_id": "...", "home_team": "巨人", "away_team": "ヤクルト",
  "start_time_jst": null, "status": "in_progress", "score": {"home": 3, "away": 2}, "url": "..." }

// Game ended
{ "game_id": "...", "home_team": "ロッテ", "away_team": "西武",
  "start_time_jst": null, "status": "final", "score": {"home": 5, "away": 1}, "url": "..." }

// No games scheduled (off-day)
{ "scoreboard": { "sport": "npb_baseball", "date_label": null, "games": [] } }
Yahoo! JAPAN Homepage Briefing · browse.sh