Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate a detailed, professional video content summary from timestamped subtitles/transcripts (e.g., lines starting with 00:00 / 1:23:45). Enforce strict per-segment structure (timestamp range + bold segment title + 2-paragraph body: first-person creator summary + expert 【导师评注】 critique with uncertainty handling). Use when the user provides time-coded subtitles and asks for a规范化纪要/内容纪要/逐段总结, and optionally wants a clean PDF export (do NOT include the full raw transcript in the PDF unless explic
.claude/skills/cnfjlhj-timestamped-video-summary/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 22% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 22% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 5% | 0% |
将“带时间戳的字幕/转录”整理为严格结构的视频内容纪要，并在用户要求“落盘/导出/PDF”时，把纪要渲染成排版清晰的中文 PDF（默认不附带原始逐行字幕材料）。
1) 只要文字纪要
2) 需要落盘为 PDF（排版清楚）
*.md → 运行 scripts/validate_summary_md.py 校验 → 运行 scripts/render_pdf.py 生成 *.pdf。00:00 / 00:02 / 01:23 / 1:23:45 这样的时间戳行对字幕的每个逻辑段落，输出必须包含且仅包含以下三部分（按顺序）：
1) 时间戳范围
时间戳范围: [开始时间 - 结束时间]2) 段落核心标题
**标题** 3) 内容主体（两层，但不加额外小标题/前缀）
【导师评注】 开头；以顶尖专家口吻补充概念、指出漏洞/争议；若无法判断真伪，必须明确写出该点“需要进一步验证”，并给出具体验证思路/方法。禁止项（默认规则）：
【导师评注】。将字幕分成“逻辑段落”时：
【导师评注】 开头）落盘流程建议： 1) 先把最终纪要写入 output.md 2) 运行校验：python3 scripts/validate_summary_md.py output.md 3) 生成 PDF：python3 scripts/render_pdf.py --input output.md --outdir .
命名规则（默认不覆盖）：
视频纪要_<主题短名>_<YYYYMMDD_HHMMSS>.pdf-v2/-v3scripts/validate_summary_md.py：格式与硬性规范校验scripts/render_pdf.py：把纪要 Markdown 渲染为 PDF（封面+目录+模块化段落；默认不附原字幕）| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-13 | fail→pass | 14,368 | 3,834 | -73% | 1 | 1 | 0% | 2,092 | 1,644 | -21% | 0 | 0 | — |
case-01 | fail→pass | 9,438 | 3,975 | -58% | 1 | 1 | 0% | 1,376 | 1,681 | +22% | 0 | 0 | — |
case-02 | fail→fail | 12,683 | 4,917 | -61% | 1 | 1 | 0% | 2,003 | 1,936 | -3% | 0 | 0 | — |
case-03 | fail→fail | 17,281 | 5,635 | -67% | 1 | 1 | 0% | 2,347 | 1,975 | -16% | 0 | 0 | — |
case-04 | fail→pass | 13,596 | 3,426 | -75% | 1 | 1 | 0% | 2,303 | 1,566 | -32% | 0 | 0 | — |
case-05 | fail→pass | 9,860 | 4,562 | -54% | 1 | 1 | 0% | 1,476 | 1,800 | +22% | 0 | 0 | — |
case-06 | fail→pass | 16,467 | 10,783 | -35% | 1 | 1 | 0% | 2,469 | 2,582 | +5% | 0 | 0 | — |
case-07 | fail→pass | 15,794 | 5,932 | -62% | 1 | 1 | 0% | 2,203 | 1,883 | -15% | 0 | 0 | — |
case-08 | fail→pass | 15,160 | 6,364 | -58% | 1 | 1 | 0% | 2,214 | 2,009 | -9% | 0 | 0 | — |
case-09 | fail→pass | 17,444 | 8,084 | -54% | 1 | 1 | 0% | 2,370 | 2,205 | -7% | 0 | 0 | — |
case-10 | pass→pass | 15,992 | 6,734 | -58% | 1 | 1 | 0% | 2,310 | 1,963 | -15% | 0 | 0 | — |
case-11 | fail→pass | 8,802 | 1,800 | -80% | 1 | 1 | 0% | 1,321 | 1,279 | -3% | 0 | 0 | — |
case-12 | fail→pass | 9,292 | 3,029 | -67% | 1 | 1 | 0% | 1,493 | 1,479 | -1% | 0 | 0 | — |
case-14 | pass→fail | 14,320 | 7,695 | -46% | 1 | 1 | 0% | 2,136 | 2,327 | +9% | 0 | 0 | — |
case-15 | pass→pass | 14,685 | 8,417 | -43% | 1 | 1 | 0% | 2,044 | 2,370 | +16% | 0 | 0 | — |
case-16 | fail→pass | 13,129 | 4,588 | -65% | 1 | 1 | 0% | 2,238 | 1,690 | -24% | 0 | 0 | — |
case-17 | pass→pass | 16,021 | 7,720 | -52% | 1 | 1 | 0% | 2,383 | 2,178 | -9% | 0 | 0 | — |
case-18 | fail→pass | 11,407 | 6,152 | -46% | 1 | 1 | 0% | 1,613 | 2,086 | +29% | 0 | 0 | — |
case-19 | fail→pass | 18,493 | 11,223 | -39% | 1 | 1 | 0% | 2,856 | 2,996 | +5% | 0 | 0 | — |
case-20 | pass→pass | 16,010 | 7,753 | -52% | 1 | 1 | 0% | 2,734 | 2,318 | -15% | 0 | 0 | — |
case-21 | pass→pass | 9,927 | 9,715 | -2% | 1 | 1 | 0% | 1,398 | 2,429 | +74% | 0 | 0 | — |
case-22 | fail→fail | 5,862 | 3,089 | -47% | 1 | 1 | 0% | 905 | 1,525 | +69% | 0 | 0 | — |
case-23 | fail→fail | 4,072 | 4,144 | +2% | 1 | 1 | 0% | 613 | 1,644 | +168% | 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. The headline lift of +52 percentage points is the difference between those two pass rates over the 23 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.