Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the spoken content of a YouTube video is needed — even if not explicitly requested: pasted video links or IDs, requests to summarize, quote, transcribe, translate, fact-check, or extract anything from a video. Also use for research or learning when a video is the source. Not for uploads or account management.
.claude/skills/bilal140202-transcript/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-10 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-20 | ✗→✓ | ▲ Improved | -18% | 0% |
Fetch video transcripts via TranscriptAPI.com.
If $TRANSCRIPT_API_KEY is not set, read references/auth-setup.md and follow the instructions there to get and store the key.
Every request needs two headers:
Bearer $TRANSCRIPT_API_KEYHermesAgent/0.11.0, ClaudeCode/1.0). Version is optional — agent name alone is fine. Do not omit this header or send a bare default — Cloudflare will return a 403 (error code 1010) and block the request.bashcurl -s "https://transcriptapi.com/api/v2/youtube/transcript\ ?video_url=VIDEO_URL&format=text&include_timestamp=true&send_metadata=true" \ -H "Authorization: Bearer $TRANSCRIPT_API_KEY" \ -H "User-Agent: YourAgent/1.0"
| Param | Required | Default | Values | | ------------------- | -------- | ------- | ------------------------------- | | video_url | yes | — | YouTube URL or 11-char video ID | | format | no | json | json, text | | include_timestamp | no | true | true, false | | send_metadata | no | false | true, false |
Accepts: full URLs (youtube.com/watch?v=ID), short URLs (youtu.be/ID), shorts (youtube.com/shorts/ID), or bare video IDs.
Default: Always use format=text&include_timestamp=true&send_metadata=true unless user specifies otherwise.
Response (format=json):
json{ "video_id": "dQw4w9WgXcQ", "language": "en", "transcript": [ { "text": "We're no strangers to love", "start": 18.0, "duration": 3.5 }, { "text": "You know the rules and so do I", "start": 21.5, "duration": 2.8 } ], "metadata": { "title": "Rick Astley - Never Gonna Give You Up", "author_name": "Rick Astley", "author_url": "https://www.youtube.com/@RickAstley", "thumbnail_url": "https://i.ytimg.com/vi/dQw4w9WgXcQ/maxresdefault.jpg" } }
Response (format=text):
json{ "video_id": "dQw4w9WgXcQ", "language": "en", "transcript": "[00:00:18] We're no strangers to love\n[00:00:21] You know the rules...", "metadata": {...} }
| Code | Meaning | Action | | -------- | ---------------- | ---------------------------------------------- | | 401 | Bad API key | Check key or re-setup | | 402 | No credits | Top up at transcriptapi.com/billing | | 403/1010 | Cloudflare block | Add or fix User-Agent header | | 404 | No transcript | Video may not have captions enabled | | 408 | Timeout | Retry once after 2s | | 429 | Rate limited | Wait and retry |
format=json when you need precise timestamps for quoting specific moments.include_timestamp=false for clean text suitable for translation or analysis.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→fail | 11,884 | 6,122 | -48% | 1 | 1 | 0% | 2,479 | 1,228 | -50% | 0 | 0 | — |
case-01 | fail→fail | 12,006 | 6,339 | -47% | 1 | 1 | 0% | 2,864 | 1,220 | -57% | 0 | 0 | — |
case-02 | fail→fail | 9,289 | 6,019 | -35% | 1 | 1 | 0% | 1,843 | 1,264 | -31% | 0 | 0 | — |
case-04 | pass→pass | 11,689 | 10,485 | -10% | 1 | 1 | 0% | 2,153 | 3,073 | +43% | 0 | 0 | — |
case-05 | pass→pass | 7,619 | 4,553 | -40% | 1 | 1 | 0% | 1,346 | 1,739 | +29% | 0 | 0 | — |
case-06 | pass→pass | 10,784 | 7,067 | -34% | 1 | 1 | 0% | 1,842 | 2,154 | +17% | 0 | 0 | — |
case-07 | fail→fail | 6,476 | 4,220 | -35% | 1 | 1 | 0% | 1,046 | 1,807 | +73% | 0 | 0 | — |
case-08 | pass→pass | 5,820 | 2,199 | -62% | 1 | 1 | 0% | 1,052 | 1,326 | +26% | 0 | 0 | — |
case-09 | pass→pass | 7,878 | 3,277 | -58% | 1 | 1 | 0% | 1,333 | 1,655 | +24% | 0 | 0 | — |
case-10 | fail→pass | 10,807 | 1,721 | -84% | 1 | 1 | 0% | 1,696 | 1,263 | -26% | 0 | 0 | — |
case-11 | pass→pass | 11,295 | 3,585 | -68% | 1 | 1 | 0% | 2,165 | 1,503 | -31% | 0 | 0 | — |
case-12 | fail→pass | 9,800 | 1,342 | -86% | 1 | 1 | 0% | 1,486 | 1,192 | -20% | 0 | 0 | — |
case-13 | fail→pass | 7,619 | 1,499 | -80% | 1 | 1 | 0% | 1,195 | 1,168 | -2% | 0 | 0 | — |
case-14 | pass→pass | 9,960 | 3,126 | -69% | 1 | 1 | 0% | 1,581 | 1,318 | -17% | 0 | 0 | — |
case-15 | pass→pass | 5,467 | 2,039 | -63% | 1 | 1 | 0% | 844 | 1,301 | +54% | 0 | 0 | — |
case-16 | pass→pass | 8,189 | 2,337 | -71% | 1 | 1 | 0% | 1,433 | 1,448 | +1% | 0 | 0 | — |
case-17 | fail→pass | 8,152 | 2,879 | -65% | 1 | 1 | 0% | 1,447 | 1,561 | +8% | 0 | 0 | — |
case-18 | pass→pass | 5,592 | 1,464 | -74% | 1 | 1 | 0% | 813 | 1,181 | +45% | 0 | 0 | — |
case-19 | pass→fail | 13,083 | 3,104 | -76% | 1 | 1 | 0% | 1,936 | 1,455 | -25% | 0 | 0 | — |
case-20 | fail→pass | 9,575 | 2,046 | -79% | 1 | 1 | 0% | 1,636 | 1,335 | -18% | 0 | 0 | — |
case-21 | fail→pass | 10,197 | 1,759 | -83% | 1 | 1 | 0% | 1,632 | 1,289 | -21% | 0 | 0 | — |
case-22 | fail→pass | 6,334 | 1,202 | -81% | 1 | 1 | 0% | 1,000 | 1,156 | +16% | 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 19 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 +27 percentage points is the difference between those two pass rates over the 19 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.