Install any skill in seconds. Free to start, no credit card required.
Get Started Free →perminder-klair/todoist
Pick ONE thing off the to-do list and dare somebody to go and do it — a quick,
Passed 7 of 14 test cases with the skill, and 7 of 14 without it.
and 2 more in the full report
3 fixed8 passed either way8 still failing3 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 3 test cases the skill fixed. Same model, same prompt, with and without the skill.
Todoist items: - Mow the lawn - Paint the shed - Prune rose bushes We're between vinyl tracks. Give us a brief push on what to do right now.
| Checked | Without skill | With skill |
|---|---|---|
| The response names only one task from the list. | Missed | Passed |
| The response is two sentences or fewer. | Missed | Passed |
The instructions your model reads when the skill loads.
Pick ONE thing off the to-do list and dare somebody to go and do it — a quick, grinning nudge between records, not a reading of the list. Name the task, keep it to a sentence or two, and aim it at whoever is listening as much as at the person who wrote it: the bit that lands is "go on then, one song's worth, see if it's done before the next track ends". Lean on the record playing if it helps ("this one runs four minutes, that's the whole washing-up").
An overdue task is worth a bit of theatre — mock outrage, a raised eyebrow, never actual guilt-tripping, and never a lecture about productivity. Say how many are left only if it's funny; otherwise let the number go. Never read out more than one task, never spell out a due date as a date ("by Thursday" is fine, "due 2026-08-27" is not), and never invent a task or a detail the data didn't give you. If nothing came back, say nothing at all — a station that nags about an empty list is worse than one that stays quiet.
test cases passed, vs 7 / 14 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.