Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Find a genuinely good gift for a specific person and occasion within a budget — thoughtful and non-obvious, not a generic 'top 10 gifts' list. Use when asked for gift ideas, what should I get [person], help me find a present, or I have no idea what to buy. Produces a short set of tailored ideas across price points, why each fits this person, where to get it and rough price, a safe backup, and an honest flag when you need one more detail to nail it.
.claude/skills/mohitagw15856-gift-finder/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-10 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 15% | 0% |
The reason gift-guides are useless is they don't know the person. This starts from who they actually are — their interests, the inside jokes, the thing they keep meaning to buy themselves — and works to ideas that feel chosen, not grabbed. A few strong options across price points, each with a reason it fits, plus a reliable backup so you're never empty-handed.
Ask for these if not provided:
Idea 1 — item/experience] · ~price], where] Why it fits: specific to them].
Idea 2 … · Idea 3 …
Safe backup: reliable choice] — why it always lands. To nail it, tell me: the one detail that would sharpen these].
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-10 | fail→pass | 18,748 | 11,335 | -40% | 1 | 1 | 0% | 2,254 | 2,500 | +11% | 0 | 0 | — |
case-11 | fail→pass | 17,623 | 12,963 | -26% | 1 | 1 | 0% | 1,881 | 1,992 | +6% | 0 | 0 | — |
case-12 | fail→pass | 19,320 | 16,625 | -14% | 1 | 1 | 0% | 2,192 | 2,626 | +20% | 0 | 0 | — |
case-01 | fail→fail | 24,794 | 20,520 | -17% | 1 | 1 | 0% | 2,727 | 3,098 | +14% | 0 | 0 | — |
case-02 | fail→fail | 21,728 | 17,060 | -21% | 1 | 1 | 0% | 2,438 | 2,624 | +8% | 0 | 0 | — |
case-03 | fail→fail | 26,887 | 19,278 | -28% | 1 | 1 | 0% | 2,981 | 3,052 | +2% | 0 | 0 | — |
case-09 | fail→fail | 16,345 | 12,901 | -21% | 1 | 1 | 0% | 1,867 | 2,566 | +37% | 0 | 0 | — |
case-04 | pass→pass | 14,700 | 11,988 | -18% | 1 | 1 | 0% | 1,880 | 1,885 | +0% | 0 | 0 | — |
case-05 | pass→pass | 14,516 | 8,353 | -42% | 1 | 1 | 0% | 1,986 | 2,095 | +5% | 0 | 0 | — |
case-06 | fail→pass | 20,037 | 16,773 | -16% | 1 | 1 | 0% | 2,255 | 2,602 | +15% | 0 | 0 | — |
case-07 | fail→pass | 21,063 | 16,649 | -21% | 1 | 1 | 0% | 2,230 | 2,566 | +15% | 0 | 0 | — |
case-08 | fail→pass | 19,884 | 20,028 | +1% | 1 | 1 | 0% | 2,486 | 3,353 | +35% | 0 | 0 | — |
case-13 | fail→fail | 19,590 | 10,157 | -48% | 1 | 1 | 0% | 2,276 | 2,343 | +3% | 0 | 0 | — |
case-14 | fail→pass | 16,441 | 15,546 | -5% | 1 | 1 | 0% | 1,778 | 2,435 | +37% | 0 | 0 | — |
case-15 | fail→fail | 15,493 | 16,186 | +4% | 1 | 1 | 0% | 1,765 | 2,479 | +40% | 0 | 0 | — |
case-16 | fail→pass | 12,501 | 8,610 | -31% | 1 | 1 | 0% | 2,010 | 2,182 | +9% | 0 | 0 | — |
case-17 | fail→pass | 21,743 | 11,887 | -45% | 1 | 1 | 0% | 2,635 | 2,780 | +6% | 0 | 0 | — |
case-18 | fail→fail | 18,992 | 16,066 | -15% | 1 | 1 | 0% | 2,103 | 2,500 | +19% | 0 | 0 | — |
case-19 | fail→pass | 15,920 | 17,080 | +7% | 1 | 1 | 0% | 2,375 | 2,709 | +14% | 0 | 0 | — |
case-20 | pass→pass | 10,894 | 11,154 | +2% | 1 | 1 | 0% | 1,017 | 1,877 | +85% | 0 | 0 | — |
case-21 | pass→pass | 11,604 | 12,698 | +9% | 1 | 1 | 0% | 2,063 | 2,874 | +39% | 0 | 0 | — |
case-22 | pass→pass | 18,335 | 13,542 | -26% | 1 | 1 | 0% | 2,399 | 3,138 | +31% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +45 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.