search-bbc-worldcup

name	search-bbc-worldcup
description	Find the 5 most important FIFA World Cup 2026 stories of the day from BBC Sport's World Cup listing page and return them as JSON with verified source URLs. Load the hyperlight-sandbox skill first for sandbox/fetch_url rules.
license	MIT
allowed-tools	execute_code
metadata	{"author":"harnessagent_sandbox_demo","version":"1.0"}

Goal

Surface the 5 most important FIFA World Cup 2026 stories of the day and return them as a single JSON block.

Source policy — STRICT

The single source of truth is https://www.bbc.com/sport/football/world-cup. Do not fetch any other site. (The sandbox allow-list is BBC-only anyway.)

Required workflow

Run all steps via execute_code per the hyperlight-sandbox skill.

1. Fetch the listing page

page = call_tool("fetch_url", url="https://www.bbc.com/sport/football/world-cup")
print(page[:6000])

2. Extract candidate URLs from the LINKS: section

Every URL you ever output MUST come from this list verbatim. You are forbidden from constructing a BBC URL from a headline, slug guess, or memory. BBC slugs are opaque hashes (e.g. c1234abcdef), not human-readable phrases. If a URL is not in the LINKS: list, it does not exist as far as you are concerned.

import re
m = re.search(r"^LINKS:\n((?:  - .+\n?)+)", page, flags=re.MULTILINE)
urls = []
if m:
    for line in m.group(1).splitlines():
        line = line.strip()
        if line.startswith("- "):
            u = line[2:].strip()
            if u and u not in urls:
                urls.append(u)
print(len(urls), urls[:15])

If urls is empty, refetch the listing page once and retry. Do not proceed to step 3 until urls has at least one entry.

3. Verify each URL

Walk urls in order. For each one, in a SEPARATE execute_code block, run call_tool("fetch_url", url=u) and inspect the first line.

• If the header starts with STATUS: 200, keep it as VERIFIED. Read the story's title from the TITLE: header line and its summary from the DESCRIPTION: header line of THAT fetch.
• If it returns 4xx / 5xx / ERROR, skip it. Do not try to fix the URL — the URL is whatever the listing page said it was.

Stop once you have 5 VERIFIED URLs, or once you have tried at least 15 candidates.

4. Fallback rules

If after step 3 you still have fewer than 5 verified stories:

• You MAY fill the remaining slots from urls using the listing page's nearby anchor text + <p> snippet for title / summary. These slots do not need a 200 verification.
• The source_url in those fallback slots MUST still be a literal entry from urls. NEVER fabricate a URL.
• If urls itself has fewer than 5 entries, output as many stories as you have URLs for (could be 1–4). Do not pad with invented URLs. Returning 3 real stories is correct; returning 5 with 2 hallucinated URLs is wrong.

Hard rules on source_url

• Each source_url MUST be a verbatim entry from the LINKS: list of the step-1 listing fetch. NEVER construct a URL from a title or slug guess.
• source_url values MUST be unique across stories.
• None may equal https://www.bbc.com/sport/football/world-cup itself.
• Prefer URLs that returned STATUS: 200 in step 3.

Output format — return THIS exact JSON shape, nothing else

```json
{
  "date": "YYYY-MM-DD",
  "stories": [
    {
      "rank": 1,
      "title": "...",
      "category": "match | friendly | team_news | colour | fifa_org",
      "summary": "1–2 sentences in English",
      "source_name": "BBC Sport",
      "source_url": "https://www.bbc.com/sport/football/..."
    }
  ]
}
```

Wrap the JSON in a fenced ```json block. Do not add prose around it.