TEST-RESULTS: hook-maker
Date: 2026-10-14 · Kind: smoke test, not proof. One run (n=1), one input. No defect rate is claimed.
How: the agent that wrote the skill followed its SKILL.md step by step in the same session, on a fictional small business (a home bakery, «مخبز سنبلة»). It was not run in a fresh Claude session, so automatic triggering was not tested, and the author checking their own skill is a weaker test than an outside run.
| Run | Input | Result |
|---|---|---|
| 1 | examples/input.md: 30 s Reel, no face, Iraqi voice with a 1-sentence sample, result at second 26, real photos, no numbers | examples/expected-output.md: 5 options (one per pattern), screen lines 4-6 words, spoken lines in her dialect, each tied to the second that pays it off; a 5 × 5 check table; one recommendation and a suggested later comparison |
What worked: no numbers were invented (the input had none); every option named where the promise is paid; the check caught a soft problem in option 5 («هذا الشهر» adds time pressure the video does not justify); a factual risk in option 3 («الصبح») was pointed out instead of hidden.
What failed or is weak: word counts were done by the author, not by a tool. The dialect match rests on one sentence; it may sound off to a native speaker. No change was made to the skill after this run.
Not tested: a Carousel cover; a formal-Arabic-only request; an input with numbers and sources; a fresh Claude session or Claude.ai.