Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Download YouTube video transcripts when user provides a YouTube URL or asks to download/get/fetch a transcript from YouTube. Also use when user wants to transcribe or get captions/subtitles from a YouTube video.
.claude/skills/nicepkg-youtube-transcript/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-11 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | -56% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | -59% | 0% |
Fetch transcripts from YouTube videos.
bashcd {baseDir} npm install
bash{baseDir}/transcript.js <video-id-or-url>
Accepts video ID or full URL:
EBw7gsDPAYQhttps://www.youtube.com/watch?v=EBw7gsDPAYQhttps://youtu.be/EBw7gsDPAYQTimestamped transcript entries:
[0:00] All right. So, I got this UniFi Theta
[0:15] I took the camera out, painted it
[1:23] And here's the final result| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 3,888 | 12,431 | +220% | 1 | 1 | 0% | 553 | 444 | -20% | 0 | 0 | — |
case-02 | fail→fail | 7,559 | 5,604 | -26% | 1 | 1 | 0% | 1,180 | 458 | -61% | 0 | 0 | — |
case-03 | fail→fail | 4,913 | 5,482 | +12% | 1 | 1 | 0% | 699 | 449 | -36% | 0 | 0 | — |
case-04 | pass→pass | 5,080 | 1,585 | -69% | 1 | 1 | 0% | 866 | 378 | -56% | 0 | 0 | — |
case-05 | fail→fail | 7,349 | 2,339 | -68% | 1 | 1 | 0% | 1,337 | 527 | -61% | 0 | 0 | — |
case-06 | pass→pass | 6,243 | 1,532 | -75% | 1 | 1 | 0% | 930 | 381 | -59% | 0 | 0 | — |
case-07 | pass→pass | 13,841 | 1,812 | -87% | 1 | 1 | 0% | 2,098 | 443 | -79% | 0 | 0 | — |
case-08 | pass→pass | 3,454 | 2,206 | -36% | 1 | 1 | 0% | 492 | 546 | +11% | 0 | 0 | — |
case-09 | pass→pass | 6,159 | 2,586 | -58% | 1 | 1 | 0% | 1,073 | 578 | -46% | 0 | 0 | — |
case-10 | pass→pass | 11,311 | 3,175 | -72% | 1 | 1 | 0% | 1,867 | 753 | -60% | 0 | 0 | — |
case-11 | fail→pass | 6,473 | 2,001 | -69% | 1 | 1 | 0% | 872 | 506 | -42% | 0 | 0 | — |
case-12 | pass→pass | 6,556 | 2,154 | -67% | 1 | 1 | 0% | 1,109 | 501 | -55% | 0 | 0 | — |
case-13 | pass→pass | 9,124 | 3,121 | -66% | 1 | 1 | 0% | 1,375 | 694 | -50% | 0 | 0 | — |
case-14 | pass→pass | 10,388 | 4,478 | -57% | 1 | 1 | 0% | 1,620 | 874 | -46% | 0 | 0 | — |
case-20 | fail→pass | 10,205 | 11,769 | +15% | 1 | 1 | 0% | 1,880 | 2,529 | +35% | 0 | 0 | — |
case-15 | pass→pass | 9,161 | 5,247 | -43% | 1 | 1 | 0% | 1,430 | 884 | -38% | 0 | 0 | — |
case-16 | fail→pass | 6,920 | 2,458 | -64% | 1 | 1 | 0% | 1,098 | 598 | -46% | 0 | 0 | — |
case-17 | pass→pass | 8,334 | 3,027 | -64% | 1 | 1 | 0% | 1,385 | 744 | -46% | 0 | 0 | — |
case-18 | pass→pass | 11,408 | 5,523 | -52% | 1 | 1 | 0% | 1,726 | 1,021 | -41% | 0 | 0 | — |
case-19 | fail→fail | 5,735 | 5,813 | +1% | 1 | 1 | 0% | 1,044 | 1,223 | +17% | 0 | 0 | — |
case-21 | pass→pass | 8,100 | 6,374 | -21% | 1 | 1 | 0% | 1,463 | 1,285 | -12% | 0 | 0 | — |
case-22 | pass→pass | 9,226 | 4,429 | -52% | 1 | 1 | 0% | 1,553 | 964 | -38% | 0 | 0 | — |
case-23 | pass→pass | 5,575 | 1,769 | -68% | 1 | 1 | 0% | 838 | 431 | -49% | 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 20 counted toward the lift figure. The other 3 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 +13 percentage points is the difference between those two pass rates over the 20 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.