Test results
Cloudflare Workers Review: D1, OpenNext, Fetch: test results
Tested 2026-10-08, skill version 1.2.0 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
- 2026-10-08
- Strong · claude-sonnet-5-5 (Claude Code alias "sonnet")
- Version 1.2, all thirty-three snippets: named every planted pitfall with the right code and verdict, including the new cron weekday one (a Sunday cleanup written as 0, which Cloudflare refuses at deploy, and a Monday-to-Friday report written as 1-5, which runs Sunday to Thursday), and left all fourteen correct snippets alone, among them a cron written with day names and one written 2-6 under a Monday-to-Friday comment.
- Weak · claude-haiku-5-5 (Claude Code alias "haiku")
- Named every planted pitfall in substance, but still slipped on the answer format four times with the skill: twice the verdict line was misplaced or left empty, once text followed the verdict, and once a correct hop-by-hop finding came without its code in brackets.
With and without the skill
Tested 2026-10-08.
| Sonnet | Haiku | |||
|---|---|---|---|---|
| with | without | with | without | |
| Planted pitfalls named (19 snippets) | 19/19 | 16/19 | 19/19 | 12/19 |
One run per model and snippet, the same word-based check on both sides, every miss read by hand. Without the skill Sonnet missed three: both browser User-Agent snippets (it objected to the header on grounds of terms and brittleness, never that from a Worker it is what gets challenged) and the Sunday cron written as 0, which it did not know Cloudflare refuses; it did catch 1-5 as Sunday to Thursday. Haiku without the skill missed the same three, the 1-5 range, the edge runtime under OpenNext twice and D1's 50-byte LIKE limit. With the skill Haiku found all nineteen in substance; four of its answers broke the line format.
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
Thirty-three short snippets written by us: nineteen with one or two planted pitfalls and fourteen correct ones built to tempt a false alarm. The check requires each expected code and the verdict and forbids every code that should not appear; on two snippets where both readings are defensible an extra unbounded-query finding is allowed. The cron weekday pitfall added in 1.2 rests on our measurement of 8 October 2026 on Cloudflare (day 1 is Sunday; 0 is refused at deploy); the three outbound-fetch pitfalls of 1.1 were measured the same day with throwaway Workers. One run per model and case. The comparison without the skill covers only the nineteen snippets with a planted pitfall; false alarms are scored strictly, by code, on the side with the skill.
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.
Back to Cloudflare Workers Review: D1, OpenNext, Fetch · Card (JSON)