| name | verify-heading |
| title | Verify example.com Main Heading |
| description | >- Fetches https://example.com and verifies its main <h1> reads 'Example Domain'. Minimal end-to-end smoke test using the Browserbase Fetch API. |
| website | example.com |
| category | testing |
| tags | - testing - smoke-test - fetch-api - static-html - example |
| source | 'community: agent-runtime 2026-05-19' |
| updated | '2026-05-19' |
| recommended_method | api |
| alternative_methods | - method: browser rationale: >- Live Browserbase session works but is overkill — example.com is pure server-rendered HTML with one <h1>. Use only when the Fetch API is unavailable on the host. |
Verify example.com Main Heading
Purpose
Fetches https://example.com and verifies its main <h1> heading. Returns the heading text, the page title, and a boolean indicating whether the heading matches the canonical value Example Domain. Read-only; no auth, no forms, no JS execution required.
When to Use
- Smoke-testing a Browserbase-based stack end-to-end (API key, network egress, parsing pipeline).
- Validating a new agent's HTTP fetch path against a stable, known-good HTML response.
- Demonstrating the minimal
recommended_method: api honesty pattern (Fetch API beats live-browser for static HTML).
- Health-checking outbound connectivity in a CI/sandbox before exercising a more expensive site.
Workflow
-
Fetch the page via the Browserbase Fetch API (optimal — no browser session needed):
browse cloud fetch https://example.com --allow-redirects --output page.html
Expected response: {"ok": true, "statusCode": 200, "contentType": "text/html", "sizeBytes": ~528}.
-
Extract the first <h1> from the returned HTML. A regex is sufficient because example.com's markup is hand-written, single-line, with exactly one <h1>:
python3 -c "import re,sys; m=re.search(r'<h1[^>]*>(.*?)</h1>', open('page.html').read(), re.I|re.S); print(m.group(1).strip() if m else '')"
-
Compare against the canonical value Example Domain. If equal, return {"verified": true, ...}; otherwise return {"verified": false, "heading": "<observed>"} so the caller can investigate whether IANA changed the reference page.
-
(Optional) Also extract <title> for a secondary sanity check — it has the same value Example Domain and gives independent confirmation that the response wasn't a proxy error page.
Browser fallback
If for any reason the Fetch API is unavailable, drive a Browserbase session:
sid=$(browse cloud sessions create --keep-alive | jq -r .id)
ws=$(browse cloud sessions debug "$sid" | jq -r .wsUrl)
browse open https://example.com --cdp "$ws" --wait load
browse get text "h1" --cdp "$ws"
browse screenshot --cdp "$ws" --out final.png
browse cloud sessions update "$sid" --status REQUEST_RELEASE
Stealth/proxies are not needed — example.com is IANA's reserved demo domain with no anti-bot infrastructure.
Site-Specific Gotchas
- The page is pure server-rendered HTML — there is no JS, no XHR, no SPA hydration. Anyone reaching for a live browser to extract the H1 is over-engineering; the Fetch API returns the entire 528-byte document in one round trip.
- Exactly one
<h1>, hand-written single-line markup. A naïve regex <h1[^>]*>(.*?)</h1> works reliably; you do not need an HTML parser. Don't over-build.
- The canonical heading is
Example Domain (verified 2026-05-19 against the live IANA reference page). If you ever see something else, treat it as a signal that either (a) IANA changed the example template, or (b) you hit a captive-portal / proxy intercept page rather than the real origin.
<title> and <h1> have the same value. Two independent fields you can cross-check for free.- Redirects. Pass
--allow-redirects to browse cloud fetch defensively — at the time of authoring, https://example.com returns 200 directly with no redirect, but some networks intercept and 30x.
- CDP from restricted sandboxes. The Browserbase CDP endpoints (
connect.browserbase.com, connect.usw2.browserbase.com) are sometimes blocked even when api.browserbase.com is allowlisted. On such hosts the Fetch API path is the only viable route — another reason it's the recommended method here.
- No site-specific anti-bot caveats observed. No proxies, no stealth, no captcha, no user-agent fingerprinting. example.com is the canonical bare-friendly test domain.
Expected Output
{
"url": "https://example.com",
"status_code": 200,
"content_type": "text/html",
"size_bytes": 528,
"title": "Example Domain",
"heading": "Example Domain",
"verified": true
}
On a mismatch (defensive shape):
{
"url": "https://example.com",
"status_code": 200,
"title": "Example Domain",
"heading": "<observed-text>",
"verified": false,
"reason": "heading text differs from canonical 'Example Domain'"
}
On an upstream failure (Fetch API non-2xx, redirect loop, network block):
{
"url": "https://example.com",
"status_code": 0,
"verified": false,
"reason": "fetch failed: <error>"
}
