agentsclimarketplace

Video cut

Skill worldwonderer/video-recap-skills/skills/video-cut

把长视频按 Agent 选择的原片区间剪成短片。作为两阶段创作流程中的剪辑环节,读取 clip_plan.json 与源视频, 输出 edited_source.mp4;随后 Agent 按输出时间线写 narration.json。单独调用且未传 --no-narration-map 时, 仍支持旧版单阶段路径,把原片时间的 narration.json 映射为 narration_mapped.json。 触发词:视频剪辑、剪辑式解说、video cut、clip plan、拼剪。From its SKILL.md

Install
npx -y skills add worldwonderer/video-recap-skills --skill video-cut

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • runs commandsInstructs the agent to run 1 command, including `python3 scripts/cut.py <video> --work-dir <work_dir> [--target-duration 10m] [--clip-padding 0] [--allow-overlap]`.

SKILL.md

4.7 KB, ~1.6k tokens by cl100k_base, as published. Nobody here has run it

1. 定位

本技能只执行 Agent 已经做出的剪辑决定:

  1. 校验并补全 clip_plan.json,写出带 clip_id、原片/输出时间与时长的 clip_plan_validated.json
  2. 先避开原片硬切附近的闪帧风险,最后把边界吸附到可靠句末/自然停顿;声音完整性拥有最终优先级。
  3. 拼接选定区间,输出 edited_source.mp4
  4. 编排流程默认到此停止,由 Agent 按真实输出时间线写 narration.json
  5. 旧版单阶段路径还会把原片时间的旁白映射为 narration_mapped.json

相同输入会得到相同输出。缓存仅表示:当 edited_source.mp4 新于 clip_plan.json 时复用成片。

2. 输入契约

work_dir/clip_plan.json 可以是数组,也可以是 {"clips": [...]}

{"start": 12.0, "end": 28.5, "reason": "b02 | turn | power: A→B | POV=女主 | 保留反应 | 入点=问题落下 | 出点=沉默结束"}
  • start / end 是原片秒数;也接受 source_start / source_endin / out
  • 顶层可选 target_duration,例如 "10m"
  • 多视频项目的每个片段还必须填写 source_id
  • speech_boundary_anchors.json 与 ASR 时间段由理解阶段提供;Agent 先写大致区间,工具会尝试吸附并把仍在讲话区间内的入/出点作为 blocker 返回。

work_dir/narration.json 只在旧版单阶段路径中可选读取;该路径要求旁白使用原片时间。若允许重复或重叠片段,旁白可带 source_clip_id 消歧。

3. 剪辑意图契约

工具不会替 Agent 做创作选择。写片段前先完成本节的剪辑意图检查,并让每个区间映射到 recap_story_plan.json 的一个 beat。

使用现有自由文本 reason 保存简洁决定:

beat_id | function | change | POV | preferred moment | 入点 reason | 出点 reason

不要因为“事件重要”就保留整段;要保留最能让 change 成立的具体表演、反应、动作或揭示。理解与情绪允许时晚进早出,同时保证台词、动作和技术边界完整。

下面的 scripts/... 均相对于本技能目录。若执行器从仓库根目录启动,请给脚本路径加上本技能的绝对目录。脚本不从其他技能目录读取文件;外部输入仅限命令显式传入的视频、参数与 work_dir 产物。

4. 运行命令

python3 scripts/cut.py <video> --work-dir <work_dir> \
  [--target-duration 10m] [--clip-padding 0] [--allow-overlap]

5. 输出契约

  • clip_plan_validated.json:标准化片段,包含 clip_idsource_start/endoutput_start/endduration
  • edited_source.mp4:按计划拼接后的短视频。
  • narration_mapped.json:仅旧版单阶段路径生成;编排流程使用 --no-narration-map,不会生成该文件。

编排流程下游把 edited_source.mp4 当作视频,把 Agent 按输出时间写的 narration.json 当作旁白。

6. 边界与时间线规则

  • 旧版路径中,clip_plan.jsonnarration.json 都使用原片时间;本工具负责原片 → 输出映射。
  • 编排路径中,narration.json 直接使用剪后输出时间,不再映射。
  • 默认禁止重叠或重复原片区间;--allow-overlap 开启后,旁白应填写 source_clip_id
  • 片段起点只能位于源头、可靠句末/静音窗,或与上一片段构成无损同源连续连接;片段终点同理。ASR 判定仍在讲话且无法吸附时写入 unsafe_clip_sentence_boundary 并阻断。
  • SCENE_CUT_SNAP 默认开启:先按画面把 source start 向后、source end 向前吸附到附近硬切,随后句末吸附再做最终修正,避免视觉修正重新制造半句原声。默认范围为 SCENE_CUT_SNAP_MARGIN=0.5 秒,检测阈值为 SCENE_CUT_DETECT_THRESHOLD=0.4
  • 连续同源片段的无损连接不做句中双侧音频淡出;非连续片段仍在安全停顿内做防爆音淡入淡出。
  • 旧版旁白映射若跨越 clip 边界,不再裁短后继续:clamped_beats 永久阻断,--allow-sparse-cut 也不能绕过旁白句子完整性。

7. 能力边界

  • 不重新转写或分析视频。
  • 不写旁白,也不替 Agent 选择片段;只消费 clip_plan.json
  • 除拼接与时间映射所需处理外,不额外重编码。

What ships with it: 8 files

99.5 KB alongside SKILL.md, 8 of them executable

scripts/

Gives 0 of the 12 instructions most video audio skills give in ~1.6k tokens

Counted across 619 of the 725 authors here whose files we hold, read 2026-09-06

  • Read product marketing context firstin 13 of 619, across 7 files
  • Define the core visual thesis in one sentencein 11 of 619, across 3 files
  • Break the concept into 3 to 6 scenesin 11 of 619, across 3 files
  • Render the smallest working version firstin 11 of 619, across 3 files
  • Start with a low-quality smoke test renderin 11 of 619, across 3 files
  • Add captions for accessibility and engagementin 11 of 619, across 5 files
  • Write the scene outline before writing codein 11 of 619, across 3 files
  • Specify subject, action, camera, style, and moodin 11 of 619, across 5 files
  • Decide what each scene provesin 10 of 619, across 2 files
  • Export one clean thumbnail framein 10 of 619, across 2 files
  • Pick the right tool for the jobin 10 of 619, across 4 files
  • Run the test suite before proposing a fixin 8 of 619, across 7 files

Said here and by no other author read

  • Validate and complete clip plan JSON
  • Preserve sound integrity as top priority
  • Concatenate selected intervals into output
  • Avoid flash frame risks near hard cuts
  • Snap boundaries to reliable sentence ends
  • Check editing intent before writing clips

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.