Search TheWindowsClub for articles by keyword (with optional category/tag scoping and date/relevance sort), returning each article's title, URL, publish date, excerpt, category and tag IDs. This skill was captured from a live agent session on thewindowsclub.com and publishes here verbatim, exactly as an agent receives it.

Purpose

Return a list of TheWindowsClub articles matching a keyword query — each with title, canonical URL, publish date (local + GMT), HTML excerpt, author id, category ids, and tag ids. Optionally scope the search to a category (broad bucket: Windows / General / Office / Downloads / Security) or a tag (any of 399 topical labels: Outlook, Excel, Chrome, Edge, Errors, Troubleshoot, Windows Updates, etc.). Read-only; never posts, comments, or interacts with login-gated routes.

When to Use

Daily / hourly monitoring of new TheWindowsClub posts on a Windows topic (BSOD, Windows Update errors, registry tweaks, Edge/Chrome issues, Office 365 problems).
Hydrating known article IDs into full title/excerpt/link records for downstream summarisation.
Bulk extraction of every article in a category or tag (e.g. all 715 Outlook-tagged posts, all 848 Security category posts) for offline indexing.
Anywhere you would otherwise scrape TheWindowsClub HTML — the WP REST API is faster, returns structured fields, and is Cloudflare-cached.

Workflow

TheWindowsClub is a standard WordPress site with its public REST API exposed at https://www.thewindowsclub.com/wp-json/wp/v2/... — no auth, no cookies, no anti-bot challenge, no stealth requirement. Cloudflare fronts the origin and caches responses (Cf-Cache-Status: HIT on repeat queries; max-age=691200 ≈ 8 days). Residential proxies are not required for the API (a direct HTTP fetch with default egress returns 200 OK), but most browser-sandbox environments have outbound HTTP firewalled, so route every request through a direct HTTP fetch or a Browserbase session. Lead with the API path; the browser path costs ~10-30× more turns per result and truncates excerpts.

Build the search URL. The primary endpoint is /wp-json/wp/v2/posts. Keep the response small with _fields= and tune sort:

GET https://www.thewindowsclub.com/wp-json/wp/v2/posts
    ?search={url-encoded query}
    &per_page={1..100}          # WP cap is 100; default is 10
    &page={N}                   # 1-indexed
    &orderby={date|relevance|modified|title|id}
    &order={desc|asc}            # default desc
    &categories={id}             # optional broad-bucket filter
    &tags={id}                   # optional topical-tag filter
    &_fields=id,date,date_gmt,modified,slug,link,title,excerpt,author,categories,tags

Default sort is orderby=date&order=desc (newest first). orderby=relevance is only honoured when search= is also supplied and produces materially different (better-matched) results — e.g. search=fix+windows+update with default sort returns the latest "Windows Update" article (any topic); with orderby=relevance it returns "Fix Windows Update error 0x80070BC9" at rank 1.

Read the response totals from headers, not the body — WP returns the items array only:
- X-Wp-Total — total matching posts (e.g. 728 for search=fix blue screen, 6979 for search=outlook).
- X-Wp-Totalpages — total pages at the current per_page (e.g. 73 at per_page=10, 8 at per_page=100).
- Link: <...page=N+1>; rel="next", <...page=N-1>; rel="prev" — RFC 5988 pagination links.
Decode each post. Every item in the JSON array has WordPress's standard shape; the fields you need:
- id — stable WP post id (e.g. 107739). Use for single-post hydration via GET /wp-json/wp/v2/posts/{id}.
- date — local publish time ("2026-05-05T03:29:00", no timezone suffix — the site's TZ is IST, UTC+05:30).
- date_gmt — UTC publish time ("2026-05-04T21:59:00"). Prefer this for sorting / "since" filters — date is timezone-bare.
- modified / modified_gmt — last edit timestamps (articles are routinely updated; modified > date is normal and not a republish).
- slug — URL slug ("logi-options-lets-you-control-and-personalize-logitech-devices").
- link — canonical article URL ("https://www.thewindowsclub.com/{slug}"). No date in path — flat slug-only URL pattern.
- title.rendered — HTML-entity-encoded title ("This calendar can’t be shared..."). Decode HTML entities before display.
- excerpt.rendered — opening-paragraph HTML, wrapped in <p>...</p>, occasionally truncated mid-word followed by […] or similar. Strip tags + decode entities for plain-text.
- author — numeric WP user id. Hydrate via GET /wp-json/wp/v2/users/{id} if you need the display name.
- categories — array of category ids. TheWindowsClub uses only 5 top-level categories: 569 Windows (11955 posts), 186 General (5520), 130 Office (2808), 8 Downloads (2750), 6 Security (848). Most posts have exactly one.
- tags — array of tag ids. 399 tags total — this is the meaningful topical taxonomy. Top tags: 11 Games, 14 Freeware, 73 Troubleshoot, 753 Errors, 424 Outlook, 435 Excel, 174 Chrome, 1176 Edge, 4 Features, 150 Windows Updates.
Construct human-readable category/tag names (optional, recommended for output). The full taxonomy fits in one request each:
```
GET /wp-json/wp/v2/categories?per_page=100&_fields=id,name,slug,count
GET /wp-json/wp/v2/tags?per_page=100&orderby=count&order=desc&_fields=id,name,slug,count
```
Categories endpoint returns only 5 items total. Tags endpoint paginates (X-Wp-Totalpages: 4 at per_page=100). Cache locally — the taxonomy changes rarely.
Paginate. Increment page= until you have enough results or reach X-Wp-Totalpages. WP returns HTTP 400 (rest_post_invalid_page_number) if you exceed total pages; stop one short.
Sub-100 batches for unbounded crawls. WP caps per_page at 100. For large result sets (e.g. all 6,979 "outlook" matches), iterate page=1..70 at per_page=100. Throttle to ~1 req/s — Cloudflare caches GETs so repeats are nearly free, but bursts on uncached queries can trip rate-limit middleware.

Lightweight alternative — the search endpoint

/wp-json/wp/v2/search returns the same set with only id, title, url, type, subtype per item (~10× smaller payload). It also includes WP pages (subtype=page), not just posts — pass subtype=post to filter. Use when you only need title + link and don't care about date/excerpt:

GET /wp-json/wp/v2/search?search={q}&subtype=post&per_page=100&page={N}

Browser fallback

If the JSON API ever returns a Cloudflare interstitial or /wp-json/ is disabled, fall back to the site's built-in search at https://www.thewindowsclub.com/?s={url-encoded query}. The page is server-rendered (snapshot returns refs; no need to wait for JS). Article cards live under repeating blocks; for each block extract:

URL: <h2 class="entry-title"><a href="(https://www\.thewindowsclub\.com/[^"]+)"
Title: text content of the same <a> (HTML-entity decoded)
Date: <time[^>]+datetime="([^"]+)" (ISO 8601, IST)
Excerpt: <div class="entry-summary">\s*<p>([^<]+)</p> (truncated to ~30 words by the theme — shorter than the API's excerpt)
Author: <a rel="author"[^>]*>([^<]+)</a>

Pagination at the bottom: /page/{N}/?s={q} — same ?s= query carried forward. Capture a an html read of the body per page and run the above regex set; do not use a snapshot + click to enumerate (~3 turns per card vs. one fetch for the whole page). A Browserbase session with a stealth + residential-proxy session is recommended for the browser path because Cloudflare's bot challenge can fire on bare egress.

Site-Specific Gotchas

Cloudflare caches API GETs aggressively (max-age=691200 ≈ 8 days, Cf-Cache-Status: HIT on repeats). Identical queries return identical bytes — a freshly published article may not appear in search= results for several hours after publish if a popular query is sitting on a cached miss. For monitoring, use orderby=date&search= on each poll and de-dupe by id client-side; do not rely on X-Wp-Total changing in real-time.
date field is timezone-naive (IST = UTC+05:30) — "2026-05-05T03:29:00" is IST, not UTC. For absolute timestamps, use date_gmt which is correctly suffixed (also lacks Z but is GMT by name). Same applies to modified vs modified_gmt.
Title and excerpt are HTML-encoded. Smart quotes appear as ’, ampersands as &, etc. Always decode HTML entities before display. excerpt.rendered is wrapped in <p>...</p> — strip tags first.
Only 5 top-level categories — topical filtering lives in tags. categories=569 (Windows) covers ~12k posts and isn't a useful narrowing filter. Use tags={tag-id} (e.g. tags=424 for Outlook → 283 results when combined with search=error) for meaningful scope. Fetch the full tag list once and cache locally.
orderby=relevance is silently ignored without search= — you'll get date-desc results. Always pair orderby=relevance with a non-empty search query.
search does fuzzy multi-token AND-matching — search=fix+windows+update matches posts containing all three tokens anywhere in title/content/excerpt. There is no quoted-phrase operator; search="fix windows update" is treated the same as the unquoted version. For exact-phrase matching, post-filter the JSON by title.rendered.toLowerCase().includes(phrase).
per_page is hard-capped at 100. Requesting per_page=200 silently caps to 100 (no error). Total result count comes from headers (X-Wp-Total), not from counting items.
Page-overflow returns HTTP 400, not 200 with empty array. Requesting page=N+1 past X-Wp-Totalpages returns {"code":"rest_post_invalid_page_number","data":{"status":400}}. Check X-Wp-Totalpages and stop one short.
Excerpts are sometimes truncated mid-word with […] or …. They are not full article bodies — for full text, fetch content.rendered by omitting _fields= from the request (response will be 5-20× larger per post).
Article URL pattern is flat slug — https://www.thewindowsclub.com/{slug}, no /year/month/ prefix. Easy to construct from slug alone.
Modified timestamp ≠ republish. Articles are routinely edited (typo fixes, link refreshes). modified_gmt > date_gmt by months or years is normal; do not interpret it as a fresh publish event.
X-Robots-Tag: noindex on the /wp-json/ API responses is meta-information about the API endpoint itself (not the underlying posts) — it tells search engines not to index the API URLs. Safe to ignore for scraping.
The site exposes a sitemap at https://www.thewindowsclub.com/sitemap_index.xml (referenced in /robots.txt). For complete-archive enumeration (~25k posts), the sitemap is faster than paginating wp/v2/posts — but it has only URLs + lastmod, no titles/excerpts. Use for URL inventory; use the API for content.
Article-page a goto may report a waitForMainLoadState timeout because of slow third-party ad/analytics scripts on the article body. The DOM is interactive long before load fires — the screenshot and HTML are valid even when the navigation call returns with a timeout error. For the API path this is irrelevant; for the browser fallback, use an html read of the body rather than waiting for load.

Expected Output

{
  "query": "fix blue screen",
  "filters": {
    "categories": null,
    "tags": null,
    "orderby": "relevance",
    "order": "desc"
  },
  "total_results": 728,
  "total_pages": 8,
  "per_page": 100,
  "page": 1,
  "articles": [
    {
      "id": 107739,
      "title": "How to fix Blue Screen in Windows 11 or Windows 10",
      "slug": "blue-screen-death-windows-10",
      "url": "https://www.thewindowsclub.com/blue-screen-death-windows-10",
      "date_local": "2025-01-04T21:09:00",
      "date_gmt": "2025-01-04T15:39:00",
      "modified_gmt": "2026-01-12T10:22:00",
      "excerpt": "Windows 11/10 too has the Blue Screen of Death (BSOD) or Stop Error screen that appears when you are in the middle of something, upgrading the operating system...",
      "author_id": 136,
      "category_ids": [569],
      "category_names": ["Windows"],
      "tag_ids": [239],
      "tag_names": ["Blue Screen"]
    },
    {
      "id": 534689,
      "title": "Logi Options+ lets you control and personalize Logitech devices",
      "slug": "logi-options-lets-you-control-and-personalize-logitech-devices",
      "url": "https://www.thewindowsclub.com/logi-options-lets-you-control-and-personalize-logitech-devices",
      "date_local": "2026-05-05T03:29:00",
      "date_gmt": "2026-05-04T21:59:00",
      "modified_gmt": "2026-05-05T08:37:38",
      "excerpt": "Logitech devices are designed not just to work, but to work smarter, with added customization, comfort, and productivity-focused features...",
      "author_id": 136,
      "category_ids": [8],
      "category_names": ["Downloads"],
      "tag_ids": [14],
      "tag_names": ["Freeware"]
    }
  ]
}

Minimal-shape output when callers only need title + URL (using /wp-json/wp/v2/search):

{
  "query": "fix blue screen",
  "total_results": 729,
  "articles": [
    {
      "id": 107739,
      "title": "How to fix Blue Screen in Windows 11 or Windows 10",
      "url": "https://www.thewindowsclub.com/blue-screen-death-windows-10",
      "type": "post"
    }
  ]
}

Empty-result shape (valid query, no matches):

{
  "query": "completely-nonsense-query-xyz-zzz",
  "total_results": 0,
  "total_pages": 0,
  "articles": []
}

Page-overflow error shape (when caller paginates past total_pages):

{
  "error": "rest_post_invalid_page_number",
  "status": 400,
  "message": "The page number requested is larger than the number of pages available."
}

The Windows Club Article Search