Genspark · crawl (on Windows)
Grade 2 · stopped at Detects
It reports a missing page as a genuine failure. But on a JavaScript-rendered page it returns only the navigation chrome and still calls it a success. And on Windows, a URL containing & produces the wrong exit code.
Fetch five different kinds of page: an English help doc, a JavaScript-rendered help page, Wikipedia, a government site, and a URL that does not exist
- Runs
- Records
- Detects
- Precise
- Proven
2 / 4pages with a body where it actually returned the body
2 URLs returned no body — one gave only menus, one failed after 21 seconds
What failed Two things. (1) The Naver help page returned 586 characters, all of it menus and footer — the article body was absent, and status still said ok. The caller believes it has the page. The government site failed after waiting 21 seconds. (2) On Windows, gsk.cmd is a batch file, so cmd.exe re-parses the arguments. Everything after & in a URL is handed on as a separate command, producing "'osType' is not recognized as an internal or external command" and an exit code of 1 — while the response itself is fine. Reproduced three times; encoding & as %26 removes it. Anything checking exit codes will count a healthy request as a failure. Since characters from the URL reach the shell, it is better not to pass untrusted URLs straight through.
How we measured One fetch per URL, recording exit code, status, body length, and elapsed time separately. For the & problem: the same URL three times, then the same URL with & encoded as %26 and a version without & at all, to isolate the cause.
Not measured We did not repeat on another day (precision), nor compare the fetched text against the live page word by word (the Proven rung).
2026-09-09 · gsk CLI, as of 2026-09-09, on Windows