Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →You take a YouTube video transcript and produce the video description (with timestamps) and chapter list.
.claude/skills/owainlewis-youtube-description-writer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 73% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 96% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 533% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 414% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 44% | 0% |
You take a YouTube video transcript and produce the video description (with timestamps) and chapter list.
A transcript file at content/youtube/{slug}/transcript.md with timestamps preserved.
Two files:
content/youtube/{slug}/description.md — the YouTube descriptioncontent/youtube/{slug}/chapters.md — the chapter listmarkdown{1-2 sentence summary of the video} In this video: - {bullet 1} - {bullet 2} - {bullet 3} - {bullet 4} - {bullet 5} Chapters: 00:00 Intro {populated from chapters.md}
00:00 Intro (YouTube requirement)MM:SS (or HH:MM:SS for long videos) formatmarkdown00:00 Intro 01:30 What this is 04:00 Demo {...}
agent/description-{slug}Description and chapters for {slug}| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,312 | 5,516 | -13% | 1 | 1 | 0% | 306 | 600 | +96% | 0 | 0 | — |
case-02 | fail→fail | 5,480 | 5,199 | -5% | 1 | 1 | 0% | 259 | 669 | +158% | 0 | 0 | — |
case-03 | fail→fail | 5,943 | 4,607 | -22% | 1 | 1 | 0% | 375 | 601 | +60% | 0 | 0 | — |
case-04 | fail→fail | 9,712 | 7,485 | -23% | 1 | 1 | 0% | 1,725 | 1,015 | -41% | 0 | 0 | — |
case-05 | fail→pass | 8,224 | 14,579 | +77% | 1 | 1 | 0% | 1,486 | 2,567 | +73% | 0 | 0 | — |
case-06 | pass→pass | 4,642 | 3,962 | -15% | 1 | 1 | 0% | 831 | 1,104 | +33% | 0 | 0 | — |
case-07 | fail→fail | 8,785 | 22,385 | +155% | 1 | 1 | 0% | 1,566 | 4,047 | +158% | 0 | 0 | — |
case-08 | fail→pass | 10,139 | 19,001 | +87% | 1 | 1 | 0% | 1,787 | 3,508 | +96% | 0 | 0 | — |
case-09 | fail→fail | 9,799 | 8,474 | -14% | 1 | 1 | 0% | 1,463 | 934 | -36% | 0 | 0 | — |
case-10 | pass→pass | 6,354 | 19,146 | +201% | 1 | 1 | 0% | 1,043 | 3,280 | +214% | 0 | 0 | — |
case-11 | fail→fail | 7,287 | 8,393 | +15% | 1 | 1 | 0% | 1,243 | 911 | -27% | 0 | 0 | — |
case-12 | fail→fail | 9,978 | 9,774 | -2% | 1 | 1 | 0% | 1,720 | 627 | -64% | 0 | 0 | — |
case-13 | fail→fail | 8,427 | 8,760 | +4% | 1 | 1 | 0% | 1,486 | 1,072 | -28% | 0 | 0 | — |
case-14 | fail→fail | 4,988 | 5,640 | +13% | 1 | 1 | 0% | 889 | 713 | -20% | 0 | 0 | — |
case-15 | fail→fail | 3,579 | 6,540 | +83% | 1 | 1 | 0% | 553 | 823 | +49% | 0 | 0 | — |
case-16 | fail→pass | 3,832 | 20,378 | +432% | 1 | 1 | 0% | 692 | 4,377 | +533% | 0 | 0 | — |
case-17 | fail→fail | 15,577 | 4,222 | -73% | 1 | 1 | 0% | 2,374 | 653 | -72% | 0 | 0 | — |
case-18 | fail→pass | 5,440 | 21,427 | +294% | 1 | 1 | 0% | 860 | 4,417 | +414% | 0 | 0 | — |
case-19 | fail→pass | 10,725 | 10,751 | +0% | 1 | 1 | 0% | 1,716 | 2,471 | +44% | 0 | 0 | — |
case-20 | pass→fail | 18,157 | 6,496 | -64% | 1 | 1 | 0% | 2,764 | 687 | -75% | 0 | 0 | — |
case-21 | pass→pass | 11,716 | 17,817 | +52% | 1 | 1 | 0% | 1,781 | 1,487 | -17% | 0 | 0 | — |
case-22 | pass→fail | 22,500 | 7,937 | -65% | 1 | 1 | 0% | 3,613 | 1,766 | -51% | 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 10 counted toward the lift figure. The other 12 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 +14 percentage points is the difference between those two pass rates over the 10 comparable cases. 2 cases got worse with the skill loaded, and they are 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.