Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Turn a destination, some dates, and your vibe into a realistic day-by-day trip itinerary — paced for real humans, with a packing list and a rough budget. Use when asked to plan a trip, build a travel itinerary, what should I do in [place], or help me plan my holiday. Produces a day-by-day plan grouped by area (so you're not criss-crossing the city), must-book-ahead flags, a packing list tuned to the trip, a rough budget range, and honest notes on pace and gaps to fill with local info.
.claude/skills/mohitagw15856-trip-planner/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-22 | ✓→✗ | ▼ Worse | 42% | 0% |
| case-06 | ✓→✓ | = Same ✓ | -22% | 0% |
Most trip plans fail one of two ways: a Pinterest list with no shape, or an itinerary so packed it's a forced march. This builds a real one — clustered by neighbourhood so you're not zig-zagging, paced with actual downtime, honest about what needs booking ahead, and clear about where you should check current local info rather than trust a plan.
Ask for these if not provided:
Budget band: ~range] — swing factors: x].
Day 1 — area]: morning … · afternoon … · evening … · (built-in downtime: …) Day 2 — area]: …
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 58,141 | 36,308 | -38% | 1 | 1 | 0% | 6,728 | 5,467 | -19% | 0 | 0 | — |
case-02 | fail→pass | 43,912 | 33,640 | -23% | 1 | 1 | 0% | 5,996 | 4,674 | -22% | 0 | 0 | — |
case-03 | fail→pass | 48,698 | 33,563 | -31% | 1 | 1 | 0% | 5,959 | 4,642 | -22% | 0 | 0 | — |
case-04 | fail→fail | 58,651 | 31,142 | -47% | 1 | 1 | 0% | 6,931 | 4,272 | -38% | 0 | 0 | — |
case-05 | fail→pass | 34,943 | 35,091 | +0% | 1 | 1 | 0% | 4,424 | 4,822 | +9% | 0 | 0 | — |
case-06 | pass→pass | 42,356 | 27,743 | -35% | 1 | 1 | 0% | 4,782 | 3,746 | -22% | 0 | 0 | — |
case-07 | pass→pass | 36,982 | 36,274 | -2% | 1 | 1 | 0% | 4,780 | 4,977 | +4% | 0 | 0 | — |
case-08 | pass→pass | 27,260 | 24,571 | -10% | 1 | 1 | 0% | 3,345 | 3,693 | +10% | 0 | 0 | — |
case-09 | pass→pass | 57,159 | 24,677 | -57% | 1 | 1 | 0% | 7,274 | 4,139 | -43% | 0 | 0 | — |
case-10 | pass→pass | 44,683 | 22,973 | -49% | 1 | 1 | 0% | 5,225 | 3,569 | -32% | 0 | 0 | — |
case-11 | pass→pass | 32,103 | 27,304 | -15% | 1 | 1 | 0% | 4,463 | 4,068 | -9% | 0 | 0 | — |
case-12 | pass→pass | 39,366 | 35,781 | -9% | 1 | 1 | 0% | 5,366 | 4,763 | -11% | 0 | 0 | — |
case-13 | pass→pass | 38,713 | 25,609 | -34% | 1 | 1 | 0% | 5,283 | 4,087 | -23% | 0 | 0 | — |
case-14 | pass→pass | 52,191 | 25,611 | -51% | 1 | 1 | 0% | 6,760 | 4,159 | -38% | 0 | 0 | — |
case-15 | pass→pass | 48,914 | 22,212 | -55% | 1 | 1 | 0% | 4,457 | 3,598 | -19% | 0 | 0 | — |
case-16 | pass→pass | 30,279 | 26,525 | -12% | 1 | 1 | 0% | 3,956 | 4,125 | +4% | 0 | 0 | — |
case-17 | pass→pass | 31,653 | 25,734 | -19% | 1 | 1 | 0% | 4,185 | 3,835 | -8% | 0 | 0 | — |
case-18 | pass→pass | 43,719 | 20,897 | -52% | 1 | 1 | 0% | 6,539 | 4,839 | -26% | 0 | 0 | — |
case-19 | pass→pass | 17,531 | 25,734 | +47% | 1 | 1 | 0% | 1,878 | 2,953 | +57% | 0 | 0 | — |
case-20 | pass→pass | 11,708 | 15,120 | +29% | 1 | 1 | 0% | 2,366 | 2,734 | +16% | 0 | 0 | — |
case-21 | pass→pass | 13,066 | 14,535 | +11% | 1 | 1 | 0% | 2,216 | 3,323 | +50% | 0 | 0 | — |
case-22 | pass→fail | 24,135 | 24,253 | +0% | 1 | 1 | 0% | 2,570 | 3,659 | +42% | 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 +9 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
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.