Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Track live aircraft positions with zero API keys — adsb.lol's open ADS-B network primary, OpenSky fallback, via curl: by callsign, registration, or area. Use when asked where is this flight right now, what planes are overhead, track a tail number, or is that flight in the air. Produces the live position with altitude, speed, and heading interpreted, the overhead list for a location, and the rerunnable command — with the positions-not-schedules boundary stated honestly.
.claude/skills/mohitagw15856-flight-tracker/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 44% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 365% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 63% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 74% | 0% |
"Where's that flight right now?" has a genuinely open answer: hobbyist ADS-B receiver networks share live aircraft transponder data keylessly — adsb.lol serves it clean, OpenSky backs it up. This skill queries by callsign, hex, or area, and translates transponder math into human answers ("over Poland at 38,000 ft, ~40 minutes from Warsaw-ish airspace"). It's equally honest about the boundary: this is positions, not airline operations — gate changes, delays, and cancellations live in keyed commercial services, and pretending otherwise produces confident fiction.
Ask for these if not provided:
BAW123, often ≈ flight number with the airline's ICAO prefix — BA→BAW, LH→DLH, UA→UAL; do the mapping and say so), registration/tail number, or ICAO hexcurl -s "https://api.adsb.lol/v2/callsign/BAW123" · by registration: .../v2/reg/G-XLEB · by hex: .../v2/hex/4075a2 · overhead: .../v2/lat/51.5/lon/-0.12/dist/25 (nm). Fields per aircraft: flight (callsign), alt_baro (ft), gs (knots), track (° heading), baro_rate (ft/min climb/descent), lat/lon, t (type).curl -s "https://opensky-network.org/api/states/all?lamin=51.4&lamax=51.6&lomin=-0.4&lomax=0.2" — anonymous access, tighter rate limits; state vectors as positional arrays (index 1 callsign, 5/6 lon/lat, 7 baro alt m, 9 velocity m/s). Units differ between the two (ft/knots vs. m and m/s) — convert deliberately, never mix.baro_rate near 0 at 35,000+ ft = cruise; sustained −1,500 ft/min = descent (landing within ~30–45 min, framed as an estimate); track + position = direction in words ("heading east over Poland"). All labeled inference.The answer: "Airborne — over area] at alt] ft, speed] kn, heading direction]; descending, likely landing within ~range]." or the honest not-received line.]
| Field | Value | |---|---| Position (words + coords) · altitude · speed · heading · climb/descent · aircraft type]
Overhead mode: table sorted by distance]
Source: adsb.lol / OpenSky] community ADS-B · as of time] · rerun: [exact curl] Schedule-shaped questions: the boundary line + what position data can and can't infer]
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 9,904 | 8,501 | -14% | 1 | 1 | 0% | 2,000 | 2,885 | +44% | 0 | 0 | — |
case-02 | fail→pass | 3,096 | 9,096 | +194% | 1 | 1 | 0% | 647 | 3,010 | +365% | 0 | 0 | — |
case-03 | fail→pass | 10,305 | 10,399 | +1% | 1 | 1 | 0% | 1,923 | 3,143 | +63% | 0 | 0 | — |
case-04 | pass→pass | 9,929 | 6,761 | -32% | 1 | 1 | 0% | 2,017 | 2,532 | +26% | 0 | 0 | — |
case-05 | pass→pass | 11,233 | 8,483 | -24% | 1 | 1 | 0% | 2,169 | 2,920 | +35% | 0 | 0 | — |
case-06 | pass→pass | 10,907 | 7,936 | -27% | 1 | 1 | 0% | 2,016 | 2,741 | +36% | 0 | 0 | — |
case-07 | fail→pass | 10,605 | 5,868 | -45% | 1 | 1 | 0% | 2,254 | 2,351 | +4% | 0 | 0 | — |
case-08 | fail→pass | 8,295 | 6,521 | -21% | 1 | 1 | 0% | 1,395 | 2,427 | +74% | 0 | 0 | — |
case-09 | pass→pass | 8,333 | 6,473 | -22% | 1 | 1 | 0% | 1,505 | 2,425 | +61% | 0 | 0 | — |
case-10 | pass→pass | 10,981 | 6,832 | -38% | 1 | 1 | 0% | 1,802 | 2,419 | +34% | 0 | 0 | — |
case-11 | fail→pass | 5,695 | 6,836 | +20% | 1 | 1 | 0% | 1,013 | 2,328 | +130% | 0 | 0 | — |
case-12 | fail→pass | 12,055 | 4,923 | -59% | 1 | 1 | 0% | 2,195 | 2,167 | -1% | 0 | 0 | — |
case-13 | fail→pass | 27,895 | 4,464 | -84% | 1 | 1 | 0% | 2,483 | 2,145 | -14% | 0 | 0 | — |
case-14 | pass→pass | 8,595 | 5,084 | -41% | 1 | 1 | 0% | 1,799 | 2,309 | +28% | 0 | 0 | — |
case-15 | pass→pass | 12,109 | 7,605 | -37% | 1 | 1 | 0% | 2,109 | 2,594 | +23% | 0 | 0 | — |
case-16 | fail→fail | 14,031 | 3,530 | -75% | 1 | 1 | 0% | 2,559 | 1,925 | -25% | 0 | 0 | — |
case-17 | pass→pass | 3,666 | 6,069 | +66% | 1 | 1 | 0% | 620 | 2,405 | +288% | 0 | 0 | — |
case-18 | pass→fail | 11,361 | 4,315 | -62% | 1 | 1 | 0% | 2,053 | 2,092 | +2% | 0 | 0 | — |
case-19 | fail→pass | 6,893 | 9,094 | +32% | 1 | 1 | 0% | 1,229 | 2,798 | +128% | 0 | 0 | — |
case-20 | pass→pass | 10,304 | 4,316 | -58% | 1 | 1 | 0% | 1,782 | 2,000 | +12% | 0 | 0 | — |
case-21 | pass→pass | 6,870 | 5,856 | -15% | 1 | 1 | 0% | 1,312 | 2,266 | +73% | 0 | 0 | — |
case-22 | pass→pass | 7,554 | 6,149 | -19% | 1 | 1 | 0% | 1,498 | 2,533 | +69% | 0 | 0 | — |
case-23 | fail→pass | 6,307 | 5,444 | -14% | 1 | 1 | 0% | 1,089 | 2,124 | +95% | 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, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +39 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.