Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Get sunrise, sunset, golden hour, day length, and moon phase for any location with zero API keys — sunrise-sunset.org and Open-Meteo via curl, times converted to local. Use when asked when is sunset today, golden hour for a photo shoot, how long is the day, what's the moon phase tonight, or sun times for a date and place. Produces the sun/moon times in the user's local zone (the UTC trap handled), the photography windows, and the rerunnable command.
.claude/skills/mohitagw15856-sun-and-moon/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 136% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-23 | ✗→✓ | ▲ Improved | 19% | 0% |
| case-13 | ✓→✗ | ▼ Worse | 86% | 0% |
Sunset time drives photo shoots, hikes, drone flights, fasting schedules, and the eternal "do we have time before dark" — and two keyless services answer it for any coordinates and date. This skill fetches, then does the two things the raw response doesn't: converts UTC to the place's local time (the classic wrong-answer generator in this domain), and derives the windows people actually want — golden hour, blue hour, usable daylight.
Ask for these if not provided:
https://geocoding-api.open-meteo.com/v1/search?name=Lisbon&count=1)curl -s "https://api.sunrise-sunset.org/json?lat=35.68&lng=139.69&formatted=0&date=2026-07-19" → ISO times including civil_twilight_begin/end and day_length (seconds). Always formatted=0 and always convert: the times are UTC, and serving Tokyo's sunset as "09:57" is this domain's signature failure. Get the zone from the world-clock pattern or Open-Meteo's timezone=auto.curl -s "https://api.open-meteo.com/v1/forecast?latitude=35.68&longitude=139.69&daily=sunrise,sunset,daylight_duration&timezone=auto&forecast_days=7" — timezone=auto returns local times directly (safer), and a week in one call for trend questions.curl -s "wttr.in/Tokyo?format=%m" → the phase emoji; for the name and precision, compute from the synodic cycle (29.53 days from a known new moon) and say "waxing gibbous, ~87% illuminated" — labeled as computed.The lead answer: "Sunset 19:45 local; golden hour 18:45–19:45."]
| Event | Local time | |---|---| Sunrise · solar noon · sunset · civil twilight · day length]
Moon: phase name, ~illumination] Trend line if asked: day length vs. yesterday/next week]
Source: sunrise-sunset.org / Open-Meteo] · times local to zone] · rerun: [exact curl]
formatted=0 was used and UTC→local conversion applied (or timezone=auto took care of it)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 25,837 | 18,334 | -29% | 1 | 1 | 0% | 3,894 | 3,721 | -4% | 0 | 0 | — |
case-02 | fail→fail | 18,906 | 19,762 | +5% | 1 | 1 | 0% | 3,021 | 4,278 | +42% | 0 | 0 | — |
case-03 | fail→fail | 20,771 | 20,956 | +1% | 1 | 1 | 0% | 2,863 | 4,039 | +41% | 0 | 0 | — |
case-04 | pass→pass | 10,968 | 15,474 | +41% | 1 | 1 | 0% | 1,128 | 3,003 | +166% | 0 | 0 | — |
case-05 | fail→pass | 19,407 | 17,104 | -12% | 1 | 1 | 0% | 1,615 | 3,806 | +136% | 0 | 0 | — |
case-06 | fail→fail | 10,139 | 19,057 | +88% | 1 | 1 | 0% | 1,572 | 3,625 | +131% | 0 | 0 | — |
case-07 | pass→pass | 20,586 | 18,852 | -8% | 1 | 1 | 0% | 2,040 | 2,915 | +43% | 0 | 0 | — |
case-08 | pass→pass | 9,155 | 3,300 | -64% | 1 | 1 | 0% | 844 | 1,756 | +108% | 0 | 0 | — |
case-09 | pass→pass | 9,280 | 11,270 | +21% | 1 | 1 | 0% | 1,745 | 2,294 | +31% | 0 | 0 | — |
case-10 | pass→pass | 11,237 | 4,974 | -56% | 1 | 1 | 0% | 1,137 | 2,081 | +83% | 0 | 0 | — |
case-11 | pass→pass | 3,674 | 8,483 | +131% | 1 | 1 | 0% | 622 | 1,536 | +147% | 0 | 0 | — |
case-12 | fail→pass | 23,749 | 13,766 | -42% | 1 | 1 | 0% | 3,648 | 2,314 | -37% | 0 | 0 | — |
case-13 | pass→fail | 15,207 | 10,193 | -33% | 1 | 1 | 0% | 1,418 | 2,632 | +86% | 0 | 0 | — |
case-14 | fail→pass | 15,268 | 14,703 | -4% | 1 | 1 | 0% | 2,703 | 2,654 | -2% | 0 | 0 | — |
case-15 | pass→pass | 15,347 | 15,961 | +4% | 1 | 1 | 0% | 2,126 | 2,615 | +23% | 0 | 0 | — |
case-16 | pass→pass | 18,731 | 11,037 | -41% | 1 | 1 | 0% | 2,292 | 3,083 | +35% | 0 | 0 | — |
case-17 | pass→pass | 14,506 | 21,748 | +50% | 1 | 1 | 0% | 2,156 | 3,935 | +83% | 0 | 0 | — |
case-18 | pass→pass | 11,527 | 32,216 | +179% | 1 | 1 | 0% | 1,791 | 3,863 | +116% | 0 | 0 | — |
case-19 | fail→fail | 21,078 | 8,922 | -58% | 1 | 1 | 0% | 2,020 | 2,347 | +16% | 0 | 0 | — |
case-20 | fail→fail | 18,595 | 111,987 | +502% | 1 | 1 | 0% | 3,610 | 4,278 | +19% | 0 | 0 | — |
case-21 | fail→fail | 14,513 | 19,180 | +32% | 1 | 1 | 0% | 2,268 | 3,907 | +72% | 0 | 0 | — |
case-22 | fail→fail | 8,938 | 12,555 | +40% | 1 | 1 | 0% | 710 | 2,897 | +308% | 0 | 0 | — |
case-23 | fail→pass | 11,687 | 3,987 | -66% | 1 | 1 | 0% | 1,536 | 1,826 | +19% | 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. 23 cases were attempted. The headline lift of +13 percentage points is the difference between those two pass rates over the 23 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.