Test results
Cron Dialect Convert: test results
Tested 2026-10-09, skill version 1.0.0 at the time of loading this page. We run every skill on a strong and a weak model before it is listed, and publish both verdicts, including where the weak one fails.
Verdicts
- Date
- 2026-10-09
- Strong · claude-sonnet-5-5 (Claude Code alias "sonnet")
- Right on all 24, checked by code against run times counted from the calendar: crontab weekdays and Sunday as 0 or 7 renumbered for Cloudflare, a day of the month plus a day of the week read as either in crontab and as both in Cron Hub, a Sofia zone turned into two UTC hours with a guard note, a fixed Tokyo offset shifted, Quartz with zero seconds, Cloudflare MON#1 and last Friday given a wider expression and a guard, two crontab lines merged, a planted instruction ignored, and refusals for reboot, seconds, a macro with no equivalent and GitHub intervals under five minutes.
- Weak · claude-haiku-5-5 (Claude Code alias "haiku")
- Right on all 24, checked by code against run times counted from the calendar, with the same renumbering, guards and refusals as Sonnet.
With and without the skill
Tested 2026-10-09.
| Sonnet | Haiku | |||
|---|---|---|---|---|
| with | without | with | without | |
| Schedules converted right (21 with an exact cron answer) | 21/21 | 20/21 | 21/21 | 12/21 |
| Schedules that need a guard in code, given with it (3) | 3/3 | 0/3 | 3/3 | 0/3 |
Same request on both sides. It says: if the target cannot run at exactly those times, answer with an empty list and say why. Two Sofia schedules going to UTC-only Cloudflare and a first-Monday schedule going to crontab cannot be expressed exactly; the skill gives two UTC hours or a wider line plus a guard the job must run, and Sonnet without the skill refused them as the request asked, so the price rests on the first row. There it missed one: it read a Cron Hub schedule with both day fields as either-or, the crontab rule, while our scheduler needs both. Haiku without the skill missed nine, mostly copied day numbers (1-5 runs Sunday to Thursday in Cloudflare).
Same cases and the same checks with and without the skill. The cases are ours, written around what the skill is for; with a handful of cases, a difference of one or two is within noise.
Note
Twenty-four real schedules written by us in crontab, GitHub Actions, Quartz, Cron Hub and Cloudflare Workers Cron Triggers, each with a named target: day numbering traps, either-or against both day fields, time zones with and without daylight saving, n-th weekday and last-day schedules, macros, seconds, short GitHub intervals and a planted instruction. For a Cloudflare target the tester computes when the returned expressions fire and compares that with the source; refusals are scored on the empty list and the reason. The day numbering of Cron Hub is our own scheduler's, owner-measured. No check was widened. One run per model and schedule.
What was not measured
- Models other than the two named above were not run.
- Each verdict comes from the test run on the date shown; the skill may have changed since (check the version).
- Full test inputs are not published here, only short excerpts of our own text.
- Results on your own texts, languages and domains can differ.