Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when YouTube is or could be relevant — video/channel/playlist links, video IDs, @handles, creator lookups, summaries, quotes, topic research, tutorials, talks, product reviews, or anywhere video content is fresher or richer than text search. Covers transcripts, video/channel search, channel browsing, playlists, and in-channel search.
.claude/skills/davepoon-youtube-full/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 86% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -34% | 0% |
Complete YouTube toolkit via TranscriptAPI. Transcripts, search, channels, and playlists in one skill — no yt-dlp, no headless browser, no binary to keep patched.
Get the transcript for https://youtube.com/watch?v=VIDEO_ID and summarize it in 3 bullets.Search TED for talks about complexity theory from the last two years, then pull transcripts for the top 3 and compare their arguments.User: "Has @NASA posted anything new this week?"
Output: Calls the free channel/latest endpoint, returns new uploads with title, publish date, and view count — no credits spent.
First run prompts for a TRANSCRIPT_API_KEY (or Claude can provision one automatically). Free tier: 100 credits, no card required.
channel/latest and channel/resolve are free — use them for polling and handle lookups before spending a credit on anything else| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 8,380 | 11,456 | +37% | 1 | 1 | 0% | 1,310 | 2,436 | +86% | 0 | 0 | — |
case-02 | fail→fail | 14,164 | 7,741 | -45% | 1 | 1 | 0% | 2,195 | 739 | -66% | 0 | 0 | — |
case-03 | pass→pass | 7,711 | 2,995 | -61% | 1 | 1 | 0% | 1,472 | 1,042 | -29% | 0 | 0 | — |
case-04 | pass→pass | 8,952 | 5,027 | -44% | 1 | 1 | 0% | 1,790 | 1,356 | -24% | 0 | 0 | — |
case-05 | fail→pass | 17,780 | 10,887 | -39% | 1 | 1 | 0% | 2,764 | 2,312 | -16% | 0 | 0 | — |
case-06 | pass→pass | 10,921 | 2,115 | -81% | 1 | 1 | 0% | 1,719 | 861 | -50% | 0 | 0 | — |
case-07 | pass→pass | 9,534 | 2,479 | -74% | 1 | 1 | 0% | 1,711 | 803 | -53% | 0 | 0 | — |
case-08 | fail→pass | 4,546 | 1,900 | -58% | 1 | 1 | 0% | 851 | 818 | -4% | 0 | 0 | — |
case-09 | pass→pass | 12,932 | 7,791 | -40% | 1 | 1 | 0% | 2,420 | 1,867 | -23% | 0 | 0 | — |
case-10 | fail→pass | 10,348 | 2,466 | -76% | 1 | 1 | 0% | 1,768 | 880 | -50% | 0 | 0 | — |
case-11 | pass→pass | 7,807 | 1,945 | -75% | 1 | 1 | 0% | 1,406 | 816 | -42% | 0 | 0 | — |
case-12 | fail→pass | 13,461 | 5,643 | -58% | 1 | 1 | 0% | 2,141 | 1,416 | -34% | 0 | 0 | — |
case-13 | pass→pass | 5,911 | 3,243 | -45% | 1 | 1 | 0% | 977 | 1,004 | +3% | 0 | 0 | — |
case-14 | fail→pass | 4,720 | 1,336 | -72% | 1 | 1 | 0% | 787 | 668 | -15% | 0 | 0 | — |
case-15 | fail→pass | 11,873 | 8,428 | -29% | 1 | 1 | 0% | 2,174 | 1,881 | -13% | 0 | 0 | — |
case-16 | fail→pass | 12,146 | 6,604 | -46% | 1 | 1 | 0% | 2,277 | 1,665 | -27% | 0 | 0 | — |
case-17 | pass→pass | 17,306 | 15,515 | -10% | 1 | 1 | 0% | 3,047 | 3,199 | +5% | 0 | 0 | — |
case-18 | fail→fail | 14,540 | 6,415 | -56% | 1 | 1 | 0% | 2,447 | 1,555 | -36% | 0 | 0 | — |
case-19 | pass→pass | 12,730 | 6,673 | -48% | 1 | 1 | 0% | 2,405 | 1,772 | -26% | 0 | 0 | — |
case-20 | fail→pass | 6,358 | 6,409 | +1% | 1 | 1 | 0% | 1,201 | 1,654 | +38% | 0 | 0 | — |
case-21 | fail→pass | 5,565 | 3,257 | -41% | 1 | 1 | 0% | 1,042 | 1,020 | -2% | 0 | 0 | — |
case-22 | fail→fail | 9,658 | 15,023 | +56% | 1 | 1 | 0% | 1,735 | 3,540 | +104% | 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, and 21 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 +45 percentage points is the difference between those two pass rates over the 21 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.