Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Watch a video for the user. Use when the user shares a video URL (YouTube etc.) or local video file and wants it summarized, analyzed, or discussed — Claude can't ingest video directly, so this skill extracts scene-aware keyframes + transcript first, then reads those.
.claude/skills/huangchihhungleo-claude-real-video/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 31% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -23% | 0% |
Leo 本機裝的是 Pro 版 0.5.0（~/Projects/crv-pro/.venv/bin/crv-pro，PATH 有的環境直接 crv-pro）。
Leo 給影片（網址或檔案路徑）問內容、要摘要、拆對標 reel、問「他怎麼講/怎麼拍/觀眾為什麼買單」時。
| Leo 的話聽起來像 | 指令 | |---|---| | 「這在講什麼」「幫我摘要」「他說的是真的嗎」 | crv-pro "<src>" -o out --mode watch | | 「怎麼做的」「拆解」「為什麼會紅」「能學嗎」 | crv-pro "<src>" -o out --mode creator | | 「完整分析」或意圖不明 | crv-pro "<src>" -o out --mode full |
Leo 有具體問題就加 --why "<他的原話>"。
--senses = 五感全開：聲音事件、語氣曲線、情緒、手勢/表情、畫面標籤（首次跑下載模型 ~2.7GB，之後 86 秒影片約 40 秒）--prosody / --audio-events / --emotion / --gesture / --scene-labels--lens content|creator|both（預設 both）決定 MANIFEST 教讀的 AI 往哪個方向分析--max-frames 60；無語音影片 --no-transcribe--cookies（2026-07-05 實測可直接抓 IG reel，這才是正路）：bash python3 -c "import json; ck=json.load(open('/Users/leo/Projects/auto-post/browse-cookies/kanisleo328-threads.json')); cs=ck if isinstance(ck,list) else ck['cookies']; print('# Netscape HTTP Cookie File'); [print('\t'.join([c['domain'],'TRUE' if c['domain'].startswith('.') else 'FALSE',c.get('path','/'),'TRUE',str(int(c.get('expirationDate',2e9))),c['name'],c['value']])) for c in cs if 'instagram' in c.get('domain','')]" > /tmp/ig-cookies.txt crv-pro "<reel url>" --cookies /tmp/ig-cookies.txt --mode full 這條也失敗才退 reference_ig_post_full_read]] 的 IG 內部 API 手動下載
MANIFEST.txt — 開頭有 lens 指令區塊照著做；含幀清單、motion 區塊、perception timeline（聲音/語氣/情緒/手勢/畫面事件，各帶秒數與信心分數）、逐字稿grids/ 九宮格連續幀；要看細節才開 frames/*.jpg（讀前縮 1500px 內）perception.json — 完整感知資料（被門檻藏掉的低信心事件在這）--kb <目錄> 存進知識庫/tmp 下的）| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 17,100 | 17,650 | +3% | 1 | 1 | 0% | 1,996 | 1,603 | -20% | 0 | 0 | — |
case-07 | pass→pass | 20,783 | 10,173 | -51% | 1 | 1 | 0% | 2,948 | 2,211 | -25% | 0 | 0 | — |
case-18 | pass→pass | 11,133 | 2,567 | -77% | 1 | 1 | 0% | 1,672 | 1,394 | -17% | 0 | 0 | — |
case-02 | fail→pass | 22,737 | 22,261 | -2% | 1 | 1 | 0% | 2,627 | 2,020 | -23% | 0 | 0 | — |
case-03 | fail→fail | 19,302 | 16,084 | -17% | 1 | 1 | 0% | 2,303 | 1,440 | -37% | 0 | 0 | — |
case-04 | fail→pass | 19,954 | 10,848 | -46% | 1 | 1 | 0% | 2,591 | 2,163 | -17% | 0 | 0 | — |
case-05 | fail→pass | 18,223 | 13,970 | -23% | 1 | 1 | 0% | 2,024 | 2,647 | +31% | 0 | 0 | — |
case-06 | fail→pass | 16,823 | 13,489 | -20% | 1 | 1 | 0% | 1,777 | 2,004 | +13% | 0 | 0 | — |
case-08 | fail→pass | 23,025 | 9,490 | -59% | 1 | 1 | 0% | 2,436 | 1,864 | -23% | 0 | 0 | — |
case-09 | fail→pass | 24,071 | 8,957 | -63% | 1 | 1 | 0% | 2,649 | 1,760 | -34% | 0 | 0 | — |
case-10 | fail→pass | 18,241 | 9,160 | -50% | 1 | 1 | 0% | 2,236 | 1,610 | -28% | 0 | 0 | — |
case-11 | fail→pass | 13,450 | 3,561 | -74% | 1 | 1 | 0% | 2,370 | 1,629 | -31% | 0 | 0 | — |
case-12 | fail→pass | 19,793 | 5,560 | -72% | 1 | 1 | 0% | 2,465 | 1,976 | -20% | 0 | 0 | — |
case-19 | fail→pass | 17,853 | 7,115 | -60% | 1 | 1 | 0% | 1,669 | 1,371 | -18% | 0 | 0 | — |
case-13 | fail→pass | 7,735 | 7,837 | +1% | 1 | 1 | 0% | 1,451 | 2,203 | +52% | 0 | 0 | — |
case-14 | pass→pass | 8,951 | 5,189 | -42% | 1 | 1 | 0% | 1,512 | 1,965 | +30% | 0 | 0 | — |
case-15 | fail→pass | 26,531 | 6,692 | -75% | 1 | 1 | 0% | 1,985 | 2,209 | +11% | 0 | 0 | — |
case-16 | pass→pass | 17,141 | 5,284 | -69% | 1 | 1 | 0% | 2,622 | 1,920 | -27% | 0 | 0 | — |
case-17 | fail→pass | 8,262 | 3,176 | -62% | 1 | 1 | 0% | 1,200 | 1,594 | +33% | 0 | 0 | — |
case-20 | pass→pass | 11,241 | 37,215 | +231% | 1 | 1 | 0% | 2,338 | 2,851 | +22% | 0 | 0 | — |
case-21 | pass→pass | 23,932 | 18,228 | -24% | 1 | 1 | 0% | 3,916 | 4,999 | +28% | 0 | 0 | — |
case-22 | pass→pass | 14,168 | 8,450 | -40% | 1 | 1 | 0% | 2,061 | 2,452 | +19% | 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 20 counted toward the lift figure. The other 2 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 +59 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/26/2026 | +52% |
| gemini-3.6-flash | verified | 8/12/2026 | +42% |
Other measured skills in the registry, with their headline benchmark lift.