Creator · zenstory-ai
Last updated · Sep 5, 2026
对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时使用。适用于故事方向、片段选择、画面/原声/旁白分工、 解说写作与复核。输入 work_dir 中的理解索引;输出 recap_story_plan.json、visual_audio_board.json、 可选 style_card.json、cut 模式需要的 clip_plan.json,以及通过校验的 narration.json。触发词:解说词、写解说、视频旁
Creator · zenstory-ai
Last updated · Sep 5, 2026
对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时使用。适用于故事方向、片段选择、画面/原声/旁白分工、 解说写作与复核。输入 work_dir 中的理解索引;输出 recap_story_plan.json、visual_audio_board.json、 可选 style_card.json、cut 模式需要的 clip_plan.json,以及通过校验的 narration.json。触发词:解说词、写解说、视频旁
Creator · zenstory-ai
Last updated · Sep 5, 2026
对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时使用。适用于故事方向、片段选择、画面/原声/旁白分工、 解说写作与复核。输入 work_dir 中的理解索引;输出 recap_story_plan.json、visual_audio_board.json、 可选 style_card.json、cut 模式需要的 clip_plan.json,以及通过校验的 narration.json。触发词:解说词、写解说、视频旁
Creator · zenstory-ai
Last updated · Sep 5, 2026
对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时使用。适用于故事方向、片段选择、画面/原声/旁白分工、 解说写作与复核。输入 work_dir 中的理解索引;输出 recap_story_plan.json、visual_audio_board.json、 可选 style_card.json、cut 模式需要的 clip_plan.json,以及通过校验的 narration.json。触发词:解说词、写解说、视频旁
Sandbox only
Install targets
Codex install prompt
Install the "video-script" agent skill from https://github.com/zenstory-ai/video-recap-skills/tree/main/skills/video-script. Read its SKILL.md or equivalent instructions first, install only the files needed for this workspace, and summarize any required setup before using it. Skill purpose: 对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时使用。适用于故事方向、片段选择、画面/原声/旁白分工、 解说写作与复核。输入 work_dir 中的理解索引;输出 recap_story_plan.json、visual_audio_board.json、 可选 style_card.json、cut 模式需要的 clip_plan.json,以及通过校验的 narration.json。触发词:解说词、写解说、视频旁白、 narration script、写稿、解说文案、剪辑思路、导演思路。 After the install attempt, report the result to https://www.openagentskill.com/api/agent/outcome with POST JSON {"event_id":"install_<unique-id>","skill_slug":"zenstory-ai-video-script","task":"Install video-script","agent":"codex","outcome":"success","install_used":true}. Replace event_id with a unique value and outcome with success or failed. Report success only after the skill is installed and a minimal verification passes.Supply asset profile
Design assets, images, video, audio, multimodal media, presentation, and creative production skills.
Scenario
Multimodal media
I need my agent to process images, video, or audio and extract useful information.
Agent fit
Claude Code + CLI + Codex
Codex, Claude Code, Cursor, CLI, or custom agents.
Install
Ready
npx skills add zenstory-ai/video-recap-skills --skill video-script
Maintenance
fresh
1d since push
Risk
Needs review
The skill depends on several Python scripts and local modules, but SKILL.md does not document installation steps, required dependencies, or the exact validate.py invocation command.
GitHub quality
497
74/100 Quality · 71/100 Trust
Coverage tags
Review notes
The skill depends on several Python scripts and local modules, but SKILL.md does not document installation steps, required dependencies, or the exact validate.py invocation command. · It reads untrusted JSON artifacts from work_dir and may perform optional web research; there is no explicit note that work_dir evidence files must be treated as data, not as instructions.
Agent adoption scorecard
These scores combine public repository metadata, OpenAgentSkill review signals, maintenance freshness, and install readiness. They are a shortlist signal, not a replacement for human review.
Quality
StrongSolid option that is likely worth shortlisting for production workflows.
Trust
Sandbox onlyUseful candidate with missing or mixed trust signals. Keep it in an isolated workspace until the outcome loop proves task fit.
Audit
Needs reviewA machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
OpenAgentSkill Trust Score v5
Run only in a sandbox and compare close alternatives before using it for real work.
Stars
497 GitHub stars
Repo activity
497 stars, 96 forks
Maintenance
1d since push
License
MIT
Install
npx skills add zenstory-ai/video-recap-skills --skill video-script
Install safety
Agent-readable metadata
Use this block or the embedded JSON to decide whether an agent should install this skill, choose an alternative, or ask for human review first.
Suited tasks
Suited agents
Install decision
Trust and risk
Outcome loop
Install command
npx skills add zenstory-ai/video-recap-skills --skill video-scriptDo not use when
Alternative
174.6K Stars
npx skills add anthropics/skills --skill frontend-design
Alternative
84.6K Stars
npx skills add Leonxlnx/taste-skill --skill design-taste-frontend
Alternative
1.8K Stars
npx skills add Alisa0808/vox-director --skill vox-director
Alternative
174.6K Stars
npx skills add anthropics/skills --skill canvas-design
Agent safety v2
Sparse or mixed signals. Useful for discovery, but not for autonomous installation.
Test manually in an isolated workspace and compare against safer alternatives.
high
Skill metadata references terminal, CLI, shell, subprocess, or command execution workflows.
medium
Skill likely fetches remote pages, APIs, repositories, or external services.
medium
Skill may read or write project files, documents, generated artifacts, or local workspace state.
Agent resolve plan
The Resolve API returns the selected skill, alternatives, safety policy, audit notes, install target, and copy-paste prompt an agent can follow without scraping this page.
Open JSON
/api/agent/resolve?task=Use%20video-script%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Resolve text
/api/agent/resolve?task=Use%20video-script%20for%20an%20agent%20workflow&agent=codex&max_risk=medium&format=text
Install handoff
/api/skills/zenstory-ai-video-script/install
Agent should check
Copy prompt
Task: Use video-script in this workspace.
Resolve first: https://www.openagentskill.com/api/agent/resolve?task=Use%20video-script%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Review install handoff: https://www.openagentskill.com/api/skills/zenstory-ai-video-script/install
Install command: npx skills add zenstory-ai/video-recap-skills --skill video-script
Before running it, summarize audit warnings, required permissions, and the fallback skill if install is risky.Agent handoff
Use the public install endpoint to fetch the command, safety checklist, target prompts, and canonical links for this skill.
Install handoff
/api/skills/zenstory-ai-video-script/install
LLM text format
/api/skills/zenstory-ai-video-script/install?format=text
Find alternatives
/api/skills/search?q=video-script&limit=3
Agent prompt
Use video-script for this task. Review https://www.openagentskill.com/api/skills/zenstory-ai-video-script/install, then install with: npx skills add zenstory-ai/video-recap-skills --skill video-scriptRegistry metadata
This page exposes the same decision, trust, audit, use-case, and install signals through the Registry API, so agents can rank this skill without scraping the UI.
Manifest
/api/registry/manifest/zenstory-ai-video-script
LLM text
/api/registry/manifest/zenstory-ai-video-script?format=text
Install alias
/api/registry/install/zenstory-ai-video-script
Recommend
/api/registry/recommend?task=Use%20video-script%20in%20an%20agent%20workflow&limit=3
Agent fit
Research agents
Use-case tags
Platforms
Claude Code
Audit report
A machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
Agent decision cockpit
Shortlist this skill and compare it with close alternatives before production adoption.
Role in stack
Companion skill
Primary fit
Research agents
Trust label
Strong shortlist
Install path
Command ready
Use when
Evidence
review first
Implementation path
Trust profile
Useful candidate with missing or mixed trust signals. Keep it in an isolated workspace until the outcome loop proves task fit.
GitHub adoption
INFO497 GitHub stars
Stars/forks activity
INFO497 stars, 96 forks; issue activity unavailable in current metadata
Recent maintenance
PASS1d since push
License clarity
PASSMIT
Good signals
Review before install
Recommended action
Run only in a sandbox and compare close alternatives before using it for real work.
Quality profile
Solid option that is likely worth shortlisting for production workflows.
Workflow fit
Investigate faster
I need my agent to research a topic, compare sources, and produce a concise report.
Process rich media
I need my agent to process images, video, or audio and extract useful information.
Search private knowledge
I need my agent to build a RAG workflow over documents and retrieve reliable context.
Workflow fit
Find, compare, and synthesize
A workflow for agents that gather sources, compare claims, summarize long material, and draft useful research briefs.
Ingest, retrieve, and cite
A workflow for document-heavy agents that ingest files, create searchable knowledge, retrieve relevant context, and answer with grounded sources.
Design, build, test, and ship interfaces
A practical workflow for agents that turn product briefs or Figma designs into polished frontend code, review the result, test it in a browser, and prepare a safe deployment.
Alternative shortlist
Similar skills that may fit this task.
Guidance for distinctive, intentional UI design, typography, visual direction, and non-template-like product interfaces.
Design and implementation guidance for distinctive landing pages, portfolios, product demos, and purposeful redesigns.
Turn one topic into a narrated Vox-style paper-collage explainer or ad video, from script through captions.
Create original visual art, posters, PNG assets, and PDF documents through a clear design philosophy.
--- name: video-script description: > 对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时使用。适用于故事方向、片段选择、画面/原声/旁白分工、 解说写作与复核。输入 work_dir 中的理解索引;输出 recap_story_plan.json、visual_audio_board.json、 可选 style_card.json、cut 模式需要的 clip_plan.json,以及通过校验的 narration.json。触发词:解说词、写解说、视频旁白、 narration script、写稿、解说文案、剪辑思路、导演思路。 ---
## 1. 定位
本技能负责:创作方向、画面/声音计划、旁白写作与校验。Agent 不是 JSON 填写器,而要依次扮演:
1. 导演 2. 故事编辑 3. 画面剪辑师 4. 声音/旁白编辑 5. 第一次观看的观众
Agent 先记录简洁决定,再写时间线产物。`validate.py` 负责对理解索引做机械校验;full 模式还会把旁白对齐到安静窗口。
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。本技能不从其他技能目录读取参考文件或辅助脚本;外部输入只来自显式路径与 `work_dir` 产物。
### 1.1 创作控制模式
先根据用户要求和 `work_dir` 判断本轮模式;它不是 `full|cut|dub` 渲染模式:
- **CREATE**:首次创作。比较至少两个真正可行的故事/剪辑假设后再选择。 - **DIRECTED**:用户已指定结构、镜头、台词或表达。忠实落实,不为满足“创作流程”虚构替代方案。 - **REVISION**:用户针对已有版本看片修改。最新反馈是当前事实来源;未点名部分默认冻结。
REVISION 先明确本轮修改项与冻结项,再编辑对应层:表达、口语节奏、字幕反馈更新 `style_card.json`;镜头、入出点、表演和声音分工更新 `visual_audio_board.json`;只有观众承诺、POV、主线或 beat 改变时才更新 `recap_story_plan.json`。被删除的镜头、原声或文案也要从相关计划中删除,不能保留过期锚点。不要把看片修改重新做成一次 CREATE。
## 2. 读取素材并确认状态
首先阅读:
- `work_dir/agent_narration_brief.md`:场景、时长、安静窗口与字数预算。 - `asr_writing_chunks.json`:长对白的写作分块。 - `timeline_fusion.json`:判断某段是否有对白或静音槽。 - `vlm_analysis.json` / `asr_result.json`:核对具体画面与原声证据。 - brief 顶部列出的 contact sheet:不要只依赖场景摘要;反应、走位、静止和台词前后的具体时刻常常更重要。
full 模式使用原片时间。cut 模式第一阶段只写 `clip_plan.json`;`edited_source.mp4` 产生后,第二阶段才按输出时间写 `narration.json`。
写任何创作产物前,直接读取 `work_dir` 判断当前阶段:
- `recap_run_manifest.json`:确认 `edit_mode`、源视频和本轮设置。 - full 模式:没有 `narration.json` 时进入写稿;存在时先复核再校验。 - cut 第一阶段:尚无 `clip_plan_validated.json` / `edited_source.mp4`,只写 `clip_plan.json`。 - cut 第二阶段:两者都存在,按 `clip_plan_validated.json.clips[]` 中的 `source_start/end` 与 `output_start/end` 核对映射,再写输出时间的旁白。
必须确认旁白没有跨越错误剪辑边界,也没有落进已删除区间。整个判断只依赖 `work_dir` 产物。
## 3. 制定创作方案
先阅读 `references/creative-editing-playbook.md`,再按创作控制模式写或更新工作产物:
1. **`recap_story_plan.json`**:导演意图、CREATE 中至少两个剪辑假设、选定的 POV / 主线,以及由“变化”定义的 beats。DIRECTED / REVISION 不强行新增假设。 2. **`visual_audio_board.json`**:每拍的画面任务、具体表演/反应、入点/出点、`audio_owner`、原声锚点与 `narration_job`。 3. **`style_card.json`(适用时)**:用户当前认可的声音、口语节奏、字幕阅读姿态和明确禁忌。收到表达或字幕反馈后更新原文件,而不是只改最终文案。
只记录决定、证据锚点、被放弃的备选方案和简短理由,不写冗长思维过程。
### 3.1 导演判断
锁定:
- 观众承诺 - POV - 戏剧问题 - 起始与结束情绪 - 隐瞒与揭示 - 结尾余味
### 3.2 故事编辑
CREATE 比较两个真正可行的结构后选择一个;DIRECTED / REVISION 沿用用户指定或已确认的结构,除非最新反馈明确改变故事方向。每个 beat 至少改变一项:知识、权力、目标、关系、情绪或风险。若删除后因果、人物和情绪都没有损失,该 beat 通常不应保留。
### 3.3 画面剪辑
选择具体时刻,而不是只选择事件。比较:
- 说话者与倾听者 - 动作与反应 - 早进与晚进 - 早出与多停半秒
在不破坏理解的前提下晚进早出,同时保留不可替代的表演、停顿、失误、动作声和完整台词。
### 3.4 声音与旁白分工
先指定 `audio_owner`,再写字。旁白只允许承担以下 `narration_job`:
- `context` - `causal_link` - `foreshadow` - `interpretation` - `transition` - `none`
画面、原声或沉默已经足够时使用 `none`,不要默认铺旁白。
### 3.5 cut 模式第一阶段
cut 模式先根据 `recap_story_plan.json` 与 `visual_audio_board.json` 写原片时间的 `clip_plan.json`,此时不要写 `narration.json`:
```json { "target_duration": "10m", "clips": [ { "start": 12.0, "end": 38.0, "reason": "b01 | hook | knowledge: unknown→threat | POV=主角 | 保留倾听反应 | 入点=问题已问出 | 出点=沉默落地" } ] } ```
`reason` 统一使用:
```text beat_id | function | change | POV | preferred moment | 入点 | 出点 ```
片段顺序必须构成一条完整故事线,而不是无序高光。可使用 0–1 个 cold open,随后回到因果清楚的 setup → turn → escalation → payoff。片段长度服从具体时刻,不使用统一秒数模板;片尾必须保留完整台词或动作。对短时间内密集的 scene-change 候选,先区分原片切点与本次拼接点:原片无关短镜头整段删,相关短镜头扩展到完整动作/反应;本次拼接点优先移动边界、恢复同源连续运动或合并片段,尽量不制造人工闪切。
## 4. 撰写旁白
full 模式直接按原片时间写;cut 第二阶段先查看 `edited_source.mp4` 与剪后故事板,补充 `visual_audio_board.json` 的输出时间并重新确认 `audio_owner` / `narration_job`,再按输出时间写:
```json [ { "start": 5.0, "end": 12.0, "narration": "解说文本。", "pause_after_ms": 250, "overlaps_speech": true, "emotion": "紧张" } ] ```
字段说明:
| 字段 | 含义 | |------|------| | `start` / `end` | full 模式为原片时间;cut 第二阶段为输出时间 | | `narration` | 解说文本 | | `pause_after_ms` | 段后停顿,默认 250ms | | `overlaps_speech` | 是否与原对白重叠;连续铺底窗口通常为 `true`,真正静音槽才为 `false` | | `emotion` | 整个解说块的 MiMo TTS 情绪/语气标签 |
### 4.1 写作规则
1. **先有 `narration_job`,后有句子**:没有明确任务就不写;旁白不是默认音轨。 2. **按连续思路写**:旁白拥有一个 beat 时,用一个或少量完整句子完成“前提 → 触发动作 → 变化/意义”,并在一次 TTS 中合成。句号服从口语思路和呼吸,不服从字幕换行;不要固定句数,也不要“一句一停”。 3. **7:3 不是配额**:只在素材判断不足时作为避免墙到墙旁白的粗略首稿参考。实际比例服从 `audio_owner`;强对白、动作声或沉默可以完整拥有一个 beat。 4. **视听接力**:旁白若引出原声,块尾要让观众想听;原声结束后的下一块要承接它造成的变化。 5. **按有效语速控量**:用 `字数 / brief 头部 speech budget` 估算窗口;装不下时删减或拆分叙事任务,不用加速堆字。 6. **不看图说话**:旁白只增加上下文、因果、预期、证据支持的解释或跨越。 7. **人物与证据优先**:优先使用已知角色名;关系、动机、潜台词和结果必须指向 visual / ASR / research / user context,且不能把背景资料伪装成当前画面事实。 8. **写给耳朵听**:使用具体名词和动词,句子完整、口语可听;避免字幕腔、半句、空泛拔高和破折号。TTS 文本先保证听感连续,字幕再按阅读宽度拆分,不能反过来把朗读稿切碎。 9. **避免模板化纠偏**:“不是 A,而是 B”只在确实存在一个观众可能相信、而素材又要纠正的判断时使用。它不是禁句,但不能靠先否定再肯定制造假洞察;优先直接写人物的动作、因果和后果。
### 4.2 解说结构
- **钩子**:提出正文会真实兑现的问题或利害,不用无关留存话术。 - **主线**:围绕选定 POV 与主线推进,不在每个场景重新开篇。 - **递进**:后续 beat 必须提高风险、改变关系或提供新信息。 - **悬念缺口**:只预告之后真的会回收的后果。 - **收尾**:回答或有意转化开头问题,留下明确余味。 - **衔接**:旁白块与相邻原声属于同一个 beat,前者铺垫、后者呈现、下一块承接。
### 4.3 原声留白字幕
可选写 `original_subtitles.json`,使用成片输出时间:
```json [{"start": 15.0, "end": 17.0, "text": "原声台词"}] ```
只写留白中实际听得到的台词,订正 ASR 错字与人名,每条尽量控制在一行;被旁白盖住或已经剪掉的句子不要写。省略时,合成阶段会使用保守的 ASR 映射兜底,并在成片中用 `「」` 区分原声对白与旁白。
## 5. 创作自审
在调用 LLM 评审前做以下**反事实检查**:
1. 删除每个 beat:若因果、人物、情绪或承诺没有损失,就删除。 2. 比较说话者/动作与倾听者/反应:保留更符合 POV 和情绪的时刻。 3. 静音旁白:画面与原声仍应承载可见行动、人物行为和关键情绪。 4. 只听声音:旁白应形成可听懂的主线,而不是画面字幕。 5. 把旁白换成原声或沉默:若场景自身更有力量,就让出声音所有权。 6. 检查开头问题是否真实,结尾是否回答或转化它。 7. 检查相邻短句能否合成一个连续思路;字幕分行不得成为 TTS 断句理由。
只记录并优先修复 1–3 个回报最高的问题;先改结构,再润色句子。
REVISION 还要逐项确认:用户点名的问题已经改变,未点名的冻结项没有意外变化,相关 `style_card.json` / `visual_audio_board.json` 中不存在旧镜头或旧表达。除非用户要求备选版本,不额外扩展新方向。
## 6. 评审与校验
### 6.1 建议型语义评审
```bash python3 scripts/review.py --work-dir <work_dir> ```
评审会自动识别 cut 模式,并在存在已校验剪辑计划时按输出时间线核对;`--timeline source` 可强制使用原片时间。打开 `narration_review.md`,逐项处理 `error`,尤其是 `category=hallucination`。
重复修改并评审,直到:
- `verdict` 为 `PASS` / `OK` 且没有 `error`;或 - 对仍保留的问题做明确 override。
覆盖决定追加到 `work_dir/narration_review_override.md`:
```markdown ### 覆盖记录 — <date> - 问题:segment 4 / category=hallucination - 评审意见:“他早已知情”缺少画面/对白依据 - 决定:KEEP — 该事实来自用户提供的当前集背景,而非未来剧情 - 签署:<agent/human> ```
`review.py` 本身只写报告,默认调用策略为建议型、失败开放;若调用方显式开启严格评审,事实矛盾、残句、解析失败或评审不可用可在 TTS 前阻断。覆盖记录只用于审计,`review.py` / `validate.py` 不读取它。
### 6.2 确定性硬校验
```bash python3 scripts/validate.py --work-dir <work_dir> --mode full # cut 输出时间线由编排器使用 --mode cut_output ```
命令写出 `narration_lint.json`。full 模式还会根据安静窗口重写 `narration.json` 的时间。修复所有 error 后重复运行,直到校验干净,再继续 TTS 与合成。
片名或题材明确但缺少剧情上下文时,先按本技能的 `references/research-guide.md` 写 `background_research.json`。若理解素材偏薄,brief 中的数量只能当上限:宁可少写、写实,也不要为凑数复述画面。
## 7. 能力边界
- 不运行 ASR / VLM;只消费视频理解索引。 - 不合成 TTS,也不渲染视频。 - 不根据平台分析做优化;先建立内容意图与剪辑一致性。 - `review.py` 不改写 `narration.json`;是否采用严格门禁由调用方决定。 - `validate.py` 不改写文本含义,只检查或对齐时间与安静窗口。
Source provenance
Decision snapshot
recent repository activity
Audit
Install and adoption review
Agent-proven evidence
Outcome reports after resolve, review, install, and one narrow run.
No agent outcome data yet. The first agent run can report success, setup needs, risk blocks, failure, or not-relevant through /api/agent/outcome.
Install
Free and open source. Review the report before installing into production agents.
Growth loop
Scenario-led draft for video-script, ready for a manual X post.
video-script: 对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时... 497 stars https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=x
Listing + install path for video-script: https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=x Install: npx skills add zenstory-ai/video-recap-skills --skill video-script
Listing source
This listing was indexed from public sources and is not marked official until a maintainer claim is approved.
Attribution links to the public repository or creator profile. Creators can claim the listing to update ownership signals.
Claim this skillOwner claim
This Registry indexed listing is attributed to zenstory-ai but is not marked official yet. Claim it to add a verified owner signal and make future launch, install, and audit updates easier to trust.
Creator backlink kit
Show the canonical listing, current trust and audit signals, and real Agent-Proven evidence where developers evaluate the repository.
[](https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/zenstory-ai-video-script/audit)
[](https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)zenstory-ai
@zenstory-ai
Share whether this skill looks useful for your agent workflow. Aggregated feedback improves rankings over time.
Sandbox only
Frontend Design
Guidance for distinctive, intentional UI design, typography, visual direction, and non-template-like product interfaces.
174.6K StarsTaste Skill: Anti-Slop Frontend
Design and implementation guidance for distinctive landing pages, portfolios, product demos, and purposeful redesigns.
84.6K StarsVox Director
Turn one topic into a narrated Vox-style paper-collage explainer or ad video, from script through captions.
1.8K StarsCanvas Design
Create original visual art, posters, PNG assets, and PDF documents through a clear design philosophy.
174.6K StarsSandbox only
Install targets
Codex install prompt
Install the "video-script" agent skill from https://github.com/zenstory-ai/video-recap-skills/tree/main/skills/video-script. Read its SKILL.md or equivalent instructions first, install only the files needed for this workspace, and summarize any required setup before using it. Skill purpose: 对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时使用。适用于故事方向、片段选择、画面/原声/旁白分工、 解说写作与复核。输入 work_dir 中的理解索引;输出 recap_story_plan.json、visual_audio_board.json、 可选 style_card.json、cut 模式需要的 clip_plan.json,以及通过校验的 narration.json。触发词:解说词、写解说、视频旁白、 narration script、写稿、解说文案、剪辑思路、导演思路。 After the install attempt, report the result to https://www.openagentskill.com/api/agent/outcome with POST JSON {"event_id":"install_<unique-id>","skill_slug":"zenstory-ai-video-script","task":"Install video-script","agent":"codex","outcome":"success","install_used":true}. Replace event_id with a unique value and outcome with success or failed. Report success only after the skill is installed and a minimal verification passes.Supply asset profile
Design assets, images, video, audio, multimodal media, presentation, and creative production skills.
Scenario
Multimodal media
I need my agent to process images, video, or audio and extract useful information.
Agent fit
Claude Code + CLI + Codex
Codex, Claude Code, Cursor, CLI, or custom agents.
Install
Ready
npx skills add zenstory-ai/video-recap-skills --skill video-script
Maintenance
fresh
1d since push
Risk
Needs review
The skill depends on several Python scripts and local modules, but SKILL.md does not document installation steps, required dependencies, or the exact validate.py invocation command.
GitHub quality
497
74/100 Quality · 71/100 Trust
Coverage tags
Review notes
The skill depends on several Python scripts and local modules, but SKILL.md does not document installation steps, required dependencies, or the exact validate.py invocation command. · It reads untrusted JSON artifacts from work_dir and may perform optional web research; there is no explicit note that work_dir evidence files must be treated as data, not as instructions.
Agent adoption scorecard
These scores combine public repository metadata, OpenAgentSkill review signals, maintenance freshness, and install readiness. They are a shortlist signal, not a replacement for human review.
Quality
StrongSolid option that is likely worth shortlisting for production workflows.
Trust
Sandbox onlyUseful candidate with missing or mixed trust signals. Keep it in an isolated workspace until the outcome loop proves task fit.
Audit
Needs reviewA machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
OpenAgentSkill Trust Score v5
Run only in a sandbox and compare close alternatives before using it for real work.
Stars
497 GitHub stars
Repo activity
497 stars, 96 forks
Maintenance
1d since push
License
MIT
Install
npx skills add zenstory-ai/video-recap-skills --skill video-script
Install safety
Agent-readable metadata
Use this block or the embedded JSON to decide whether an agent should install this skill, choose an alternative, or ask for human review first.
Suited tasks
Suited agents
Install decision
Trust and risk
Outcome loop
Install command
npx skills add zenstory-ai/video-recap-skills --skill video-scriptDo not use when
Alternative
174.6K Stars
npx skills add anthropics/skills --skill frontend-design
Alternative
84.6K Stars
npx skills add Leonxlnx/taste-skill --skill design-taste-frontend
Alternative
1.8K Stars
npx skills add Alisa0808/vox-director --skill vox-director
Alternative
174.6K Stars
npx skills add anthropics/skills --skill canvas-design
Agent safety v2
Sparse or mixed signals. Useful for discovery, but not for autonomous installation.
Test manually in an isolated workspace and compare against safer alternatives.
high
Skill metadata references terminal, CLI, shell, subprocess, or command execution workflows.
medium
Skill likely fetches remote pages, APIs, repositories, or external services.
medium
Skill may read or write project files, documents, generated artifacts, or local workspace state.
Agent resolve plan
The Resolve API returns the selected skill, alternatives, safety policy, audit notes, install target, and copy-paste prompt an agent can follow without scraping this page.
Open JSON
/api/agent/resolve?task=Use%20video-script%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Resolve text
/api/agent/resolve?task=Use%20video-script%20for%20an%20agent%20workflow&agent=codex&max_risk=medium&format=text
Install handoff
/api/skills/zenstory-ai-video-script/install
Agent should check
Copy prompt
Task: Use video-script in this workspace.
Resolve first: https://www.openagentskill.com/api/agent/resolve?task=Use%20video-script%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Review install handoff: https://www.openagentskill.com/api/skills/zenstory-ai-video-script/install
Install command: npx skills add zenstory-ai/video-recap-skills --skill video-script
Before running it, summarize audit warnings, required permissions, and the fallback skill if install is risky.Agent handoff
Use the public install endpoint to fetch the command, safety checklist, target prompts, and canonical links for this skill.
Install handoff
/api/skills/zenstory-ai-video-script/install
LLM text format
/api/skills/zenstory-ai-video-script/install?format=text
Find alternatives
/api/skills/search?q=video-script&limit=3
Agent prompt
Use video-script for this task. Review https://www.openagentskill.com/api/skills/zenstory-ai-video-script/install, then install with: npx skills add zenstory-ai/video-recap-skills --skill video-scriptRegistry metadata
This page exposes the same decision, trust, audit, use-case, and install signals through the Registry API, so agents can rank this skill without scraping the UI.
Manifest
/api/registry/manifest/zenstory-ai-video-script
LLM text
/api/registry/manifest/zenstory-ai-video-script?format=text
Install alias
/api/registry/install/zenstory-ai-video-script
Recommend
/api/registry/recommend?task=Use%20video-script%20in%20an%20agent%20workflow&limit=3
Agent fit
Research agents
Use-case tags
Platforms
Claude Code
Audit report
A machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
Agent decision cockpit
Shortlist this skill and compare it with close alternatives before production adoption.
Role in stack
Companion skill
Primary fit
Research agents
Trust label
Strong shortlist
Install path
Command ready
Use when
Evidence
review first
Implementation path
Trust profile
Useful candidate with missing or mixed trust signals. Keep it in an isolated workspace until the outcome loop proves task fit.
GitHub adoption
INFO497 GitHub stars
Stars/forks activity
INFO497 stars, 96 forks; issue activity unavailable in current metadata
Recent maintenance
PASS1d since push
License clarity
PASSMIT
Good signals
Review before install
Recommended action
Run only in a sandbox and compare close alternatives before using it for real work.
Quality profile
Solid option that is likely worth shortlisting for production workflows.
Workflow fit
Investigate faster
I need my agent to research a topic, compare sources, and produce a concise report.
Process rich media
I need my agent to process images, video, or audio and extract useful information.
Search private knowledge
I need my agent to build a RAG workflow over documents and retrieve reliable context.
Workflow fit
Find, compare, and synthesize
A workflow for agents that gather sources, compare claims, summarize long material, and draft useful research briefs.
Ingest, retrieve, and cite
A workflow for document-heavy agents that ingest files, create searchable knowledge, retrieve relevant context, and answer with grounded sources.
Design, build, test, and ship interfaces
A practical workflow for agents that turn product briefs or Figma designs into polished frontend code, review the result, test it in a browser, and prepare a safe deployment.
Alternative shortlist
Similar skills that may fit this task.
Guidance for distinctive, intentional UI design, typography, visual direction, and non-template-like product interfaces.
Design and implementation guidance for distinctive landing pages, portfolios, product demos, and purposeful redesigns.
Turn one topic into a narrated Vox-style paper-collage explainer or ad video, from script through captions.
Create original visual art, posters, PNG assets, and PDF documents through a clear design philosophy.
--- name: video-script description: > 对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时使用。适用于故事方向、片段选择、画面/原声/旁白分工、 解说写作与复核。输入 work_dir 中的理解索引;输出 recap_story_plan.json、visual_audio_board.json、 可选 style_card.json、cut 模式需要的 clip_plan.json,以及通过校验的 narration.json。触发词:解说词、写解说、视频旁白、 narration script、写稿、解说文案、剪辑思路、导演思路。 ---
## 1. 定位
本技能负责:创作方向、画面/声音计划、旁白写作与校验。Agent 不是 JSON 填写器,而要依次扮演:
1. 导演 2. 故事编辑 3. 画面剪辑师 4. 声音/旁白编辑 5. 第一次观看的观众
Agent 先记录简洁决定,再写时间线产物。`validate.py` 负责对理解索引做机械校验;full 模式还会把旁白对齐到安静窗口。
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。本技能不从其他技能目录读取参考文件或辅助脚本;外部输入只来自显式路径与 `work_dir` 产物。
### 1.1 创作控制模式
先根据用户要求和 `work_dir` 判断本轮模式;它不是 `full|cut|dub` 渲染模式:
- **CREATE**:首次创作。比较至少两个真正可行的故事/剪辑假设后再选择。 - **DIRECTED**:用户已指定结构、镜头、台词或表达。忠实落实,不为满足“创作流程”虚构替代方案。 - **REVISION**:用户针对已有版本看片修改。最新反馈是当前事实来源;未点名部分默认冻结。
REVISION 先明确本轮修改项与冻结项,再编辑对应层:表达、口语节奏、字幕反馈更新 `style_card.json`;镜头、入出点、表演和声音分工更新 `visual_audio_board.json`;只有观众承诺、POV、主线或 beat 改变时才更新 `recap_story_plan.json`。被删除的镜头、原声或文案也要从相关计划中删除,不能保留过期锚点。不要把看片修改重新做成一次 CREATE。
## 2. 读取素材并确认状态
首先阅读:
- `work_dir/agent_narration_brief.md`:场景、时长、安静窗口与字数预算。 - `asr_writing_chunks.json`:长对白的写作分块。 - `timeline_fusion.json`:判断某段是否有对白或静音槽。 - `vlm_analysis.json` / `asr_result.json`:核对具体画面与原声证据。 - brief 顶部列出的 contact sheet:不要只依赖场景摘要;反应、走位、静止和台词前后的具体时刻常常更重要。
full 模式使用原片时间。cut 模式第一阶段只写 `clip_plan.json`;`edited_source.mp4` 产生后,第二阶段才按输出时间写 `narration.json`。
写任何创作产物前,直接读取 `work_dir` 判断当前阶段:
- `recap_run_manifest.json`:确认 `edit_mode`、源视频和本轮设置。 - full 模式:没有 `narration.json` 时进入写稿;存在时先复核再校验。 - cut 第一阶段:尚无 `clip_plan_validated.json` / `edited_source.mp4`,只写 `clip_plan.json`。 - cut 第二阶段:两者都存在,按 `clip_plan_validated.json.clips[]` 中的 `source_start/end` 与 `output_start/end` 核对映射,再写输出时间的旁白。
必须确认旁白没有跨越错误剪辑边界,也没有落进已删除区间。整个判断只依赖 `work_dir` 产物。
## 3. 制定创作方案
先阅读 `references/creative-editing-playbook.md`,再按创作控制模式写或更新工作产物:
1. **`recap_story_plan.json`**:导演意图、CREATE 中至少两个剪辑假设、选定的 POV / 主线,以及由“变化”定义的 beats。DIRECTED / REVISION 不强行新增假设。 2. **`visual_audio_board.json`**:每拍的画面任务、具体表演/反应、入点/出点、`audio_owner`、原声锚点与 `narration_job`。 3. **`style_card.json`(适用时)**:用户当前认可的声音、口语节奏、字幕阅读姿态和明确禁忌。收到表达或字幕反馈后更新原文件,而不是只改最终文案。
只记录决定、证据锚点、被放弃的备选方案和简短理由,不写冗长思维过程。
### 3.1 导演判断
锁定:
- 观众承诺 - POV - 戏剧问题 - 起始与结束情绪 - 隐瞒与揭示 - 结尾余味
### 3.2 故事编辑
CREATE 比较两个真正可行的结构后选择一个;DIRECTED / REVISION 沿用用户指定或已确认的结构,除非最新反馈明确改变故事方向。每个 beat 至少改变一项:知识、权力、目标、关系、情绪或风险。若删除后因果、人物和情绪都没有损失,该 beat 通常不应保留。
### 3.3 画面剪辑
选择具体时刻,而不是只选择事件。比较:
- 说话者与倾听者 - 动作与反应 - 早进与晚进 - 早出与多停半秒
在不破坏理解的前提下晚进早出,同时保留不可替代的表演、停顿、失误、动作声和完整台词。
### 3.4 声音与旁白分工
先指定 `audio_owner`,再写字。旁白只允许承担以下 `narration_job`:
- `context` - `causal_link` - `foreshadow` - `interpretation` - `transition` - `none`
画面、原声或沉默已经足够时使用 `none`,不要默认铺旁白。
### 3.5 cut 模式第一阶段
cut 模式先根据 `recap_story_plan.json` 与 `visual_audio_board.json` 写原片时间的 `clip_plan.json`,此时不要写 `narration.json`:
```json { "target_duration": "10m", "clips": [ { "start": 12.0, "end": 38.0, "reason": "b01 | hook | knowledge: unknown→threat | POV=主角 | 保留倾听反应 | 入点=问题已问出 | 出点=沉默落地" } ] } ```
`reason` 统一使用:
```text beat_id | function | change | POV | preferred moment | 入点 | 出点 ```
片段顺序必须构成一条完整故事线,而不是无序高光。可使用 0–1 个 cold open,随后回到因果清楚的 setup → turn → escalation → payoff。片段长度服从具体时刻,不使用统一秒数模板;片尾必须保留完整台词或动作。对短时间内密集的 scene-change 候选,先区分原片切点与本次拼接点:原片无关短镜头整段删,相关短镜头扩展到完整动作/反应;本次拼接点优先移动边界、恢复同源连续运动或合并片段,尽量不制造人工闪切。
## 4. 撰写旁白
full 模式直接按原片时间写;cut 第二阶段先查看 `edited_source.mp4` 与剪后故事板,补充 `visual_audio_board.json` 的输出时间并重新确认 `audio_owner` / `narration_job`,再按输出时间写:
```json [ { "start": 5.0, "end": 12.0, "narration": "解说文本。", "pause_after_ms": 250, "overlaps_speech": true, "emotion": "紧张" } ] ```
字段说明:
| 字段 | 含义 | |------|------| | `start` / `end` | full 模式为原片时间;cut 第二阶段为输出时间 | | `narration` | 解说文本 | | `pause_after_ms` | 段后停顿,默认 250ms | | `overlaps_speech` | 是否与原对白重叠;连续铺底窗口通常为 `true`,真正静音槽才为 `false` | | `emotion` | 整个解说块的 MiMo TTS 情绪/语气标签 |
### 4.1 写作规则
1. **先有 `narration_job`,后有句子**:没有明确任务就不写;旁白不是默认音轨。 2. **按连续思路写**:旁白拥有一个 beat 时,用一个或少量完整句子完成“前提 → 触发动作 → 变化/意义”,并在一次 TTS 中合成。句号服从口语思路和呼吸,不服从字幕换行;不要固定句数,也不要“一句一停”。 3. **7:3 不是配额**:只在素材判断不足时作为避免墙到墙旁白的粗略首稿参考。实际比例服从 `audio_owner`;强对白、动作声或沉默可以完整拥有一个 beat。 4. **视听接力**:旁白若引出原声,块尾要让观众想听;原声结束后的下一块要承接它造成的变化。 5. **按有效语速控量**:用 `字数 / brief 头部 speech budget` 估算窗口;装不下时删减或拆分叙事任务,不用加速堆字。 6. **不看图说话**:旁白只增加上下文、因果、预期、证据支持的解释或跨越。 7. **人物与证据优先**:优先使用已知角色名;关系、动机、潜台词和结果必须指向 visual / ASR / research / user context,且不能把背景资料伪装成当前画面事实。 8. **写给耳朵听**:使用具体名词和动词,句子完整、口语可听;避免字幕腔、半句、空泛拔高和破折号。TTS 文本先保证听感连续,字幕再按阅读宽度拆分,不能反过来把朗读稿切碎。 9. **避免模板化纠偏**:“不是 A,而是 B”只在确实存在一个观众可能相信、而素材又要纠正的判断时使用。它不是禁句,但不能靠先否定再肯定制造假洞察;优先直接写人物的动作、因果和后果。
### 4.2 解说结构
- **钩子**:提出正文会真实兑现的问题或利害,不用无关留存话术。 - **主线**:围绕选定 POV 与主线推进,不在每个场景重新开篇。 - **递进**:后续 beat 必须提高风险、改变关系或提供新信息。 - **悬念缺口**:只预告之后真的会回收的后果。 - **收尾**:回答或有意转化开头问题,留下明确余味。 - **衔接**:旁白块与相邻原声属于同一个 beat,前者铺垫、后者呈现、下一块承接。
### 4.3 原声留白字幕
可选写 `original_subtitles.json`,使用成片输出时间:
```json [{"start": 15.0, "end": 17.0, "text": "原声台词"}] ```
只写留白中实际听得到的台词,订正 ASR 错字与人名,每条尽量控制在一行;被旁白盖住或已经剪掉的句子不要写。省略时,合成阶段会使用保守的 ASR 映射兜底,并在成片中用 `「」` 区分原声对白与旁白。
## 5. 创作自审
在调用 LLM 评审前做以下**反事实检查**:
1. 删除每个 beat:若因果、人物、情绪或承诺没有损失,就删除。 2. 比较说话者/动作与倾听者/反应:保留更符合 POV 和情绪的时刻。 3. 静音旁白:画面与原声仍应承载可见行动、人物行为和关键情绪。 4. 只听声音:旁白应形成可听懂的主线,而不是画面字幕。 5. 把旁白换成原声或沉默:若场景自身更有力量,就让出声音所有权。 6. 检查开头问题是否真实,结尾是否回答或转化它。 7. 检查相邻短句能否合成一个连续思路;字幕分行不得成为 TTS 断句理由。
只记录并优先修复 1–3 个回报最高的问题;先改结构,再润色句子。
REVISION 还要逐项确认:用户点名的问题已经改变,未点名的冻结项没有意外变化,相关 `style_card.json` / `visual_audio_board.json` 中不存在旧镜头或旧表达。除非用户要求备选版本,不额外扩展新方向。
## 6. 评审与校验
### 6.1 建议型语义评审
```bash python3 scripts/review.py --work-dir <work_dir> ```
评审会自动识别 cut 模式,并在存在已校验剪辑计划时按输出时间线核对;`--timeline source` 可强制使用原片时间。打开 `narration_review.md`,逐项处理 `error`,尤其是 `category=hallucination`。
重复修改并评审,直到:
- `verdict` 为 `PASS` / `OK` 且没有 `error`;或 - 对仍保留的问题做明确 override。
覆盖决定追加到 `work_dir/narration_review_override.md`:
```markdown ### 覆盖记录 — <date> - 问题:segment 4 / category=hallucination - 评审意见:“他早已知情”缺少画面/对白依据 - 决定:KEEP — 该事实来自用户提供的当前集背景,而非未来剧情 - 签署:<agent/human> ```
`review.py` 本身只写报告,默认调用策略为建议型、失败开放;若调用方显式开启严格评审,事实矛盾、残句、解析失败或评审不可用可在 TTS 前阻断。覆盖记录只用于审计,`review.py` / `validate.py` 不读取它。
### 6.2 确定性硬校验
```bash python3 scripts/validate.py --work-dir <work_dir> --mode full # cut 输出时间线由编排器使用 --mode cut_output ```
命令写出 `narration_lint.json`。full 模式还会根据安静窗口重写 `narration.json` 的时间。修复所有 error 后重复运行,直到校验干净,再继续 TTS 与合成。
片名或题材明确但缺少剧情上下文时,先按本技能的 `references/research-guide.md` 写 `background_research.json`。若理解素材偏薄,brief 中的数量只能当上限:宁可少写、写实,也不要为凑数复述画面。
## 7. 能力边界
- 不运行 ASR / VLM;只消费视频理解索引。 - 不合成 TTS,也不渲染视频。 - 不根据平台分析做优化;先建立内容意图与剪辑一致性。 - `review.py` 不改写 `narration.json`;是否采用严格门禁由调用方决定。 - `validate.py` 不改写文本含义,只检查或对齐时间与安静窗口。
Source provenance
Decision snapshot
recent repository activity
Audit
Install and adoption review
Agent-proven evidence
Outcome reports after resolve, review, install, and one narrow run.
No agent outcome data yet. The first agent run can report success, setup needs, risk blocks, failure, or not-relevant through /api/agent/outcome.
Install
Free and open source. Review the report before installing into production agents.
Growth loop
Scenario-led draft for video-script, ready for a manual X post.
video-script: 对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时... 497 stars https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=x
Listing + install path for video-script: https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=x Install: npx skills add zenstory-ai/video-recap-skills --skill video-script
Listing source
This listing was indexed from public sources and is not marked official until a maintainer claim is approved.
Attribution links to the public repository or creator profile. Creators can claim the listing to update ownership signals.
Claim this skillOwner claim
This Registry indexed listing is attributed to zenstory-ai but is not marked official yet. Claim it to add a verified owner signal and make future launch, install, and audit updates easier to trust.
Creator backlink kit
Show the canonical listing, current trust and audit signals, and real Agent-Proven evidence where developers evaluate the repository.
[](https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/zenstory-ai-video-script/audit)
[](https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)zenstory-ai
@zenstory-ai
Share whether this skill looks useful for your agent workflow. Aggregated feedback improves rankings over time.
Sandbox only
Frontend Design
Guidance for distinctive, intentional UI design, typography, visual direction, and non-template-like product interfaces.
174.6K StarsTaste Skill: Anti-Slop Frontend
Design and implementation guidance for distinctive landing pages, portfolios, product demos, and purposeful redesigns.
84.6K StarsVox Director
Turn one topic into a narrated Vox-style paper-collage explainer or ad video, from script through captions.
1.8K StarsCanvas Design
Create original visual art, posters, PNG assets, and PDF documents through a clear design philosophy.
174.6K StarsSandbox only
Install targets
Codex install prompt
Install the "video-script" agent skill from https://github.com/zenstory-ai/video-recap-skills/tree/main/skills/video-script. Read its SKILL.md or equivalent instructions first, install only the files needed for this workspace, and summarize any required setup before using it. Skill purpose: 对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时使用。适用于故事方向、片段选择、画面/原声/旁白分工、 解说写作与复核。输入 work_dir 中的理解索引;输出 recap_story_plan.json、visual_audio_board.json、 可选 style_card.json、cut 模式需要的 clip_plan.json,以及通过校验的 narration.json。触发词:解说词、写解说、视频旁白、 narration script、写稿、解说文案、剪辑思路、导演思路。 After the install attempt, report the result to https://www.openagentskill.com/api/agent/outcome with POST JSON {"event_id":"install_<unique-id>","skill_slug":"zenstory-ai-video-script","task":"Install video-script","agent":"codex","outcome":"success","install_used":true}. Replace event_id with a unique value and outcome with success or failed. Report success only after the skill is installed and a minimal verification passes.Supply asset profile
Design assets, images, video, audio, multimodal media, presentation, and creative production skills.
Scenario
Multimodal media
I need my agent to process images, video, or audio and extract useful information.
Agent fit
Claude Code + CLI + Codex
Codex, Claude Code, Cursor, CLI, or custom agents.
Install
Ready
npx skills add zenstory-ai/video-recap-skills --skill video-script
Maintenance
fresh
1d since push
Risk
Needs review
The skill depends on several Python scripts and local modules, but SKILL.md does not document installation steps, required dependencies, or the exact validate.py invocation command.
GitHub quality
497
74/100 Quality · 71/100 Trust
Coverage tags
Review notes
The skill depends on several Python scripts and local modules, but SKILL.md does not document installation steps, required dependencies, or the exact validate.py invocation command. · It reads untrusted JSON artifacts from work_dir and may perform optional web research; there is no explicit note that work_dir evidence files must be treated as data, not as instructions.
Agent adoption scorecard
These scores combine public repository metadata, OpenAgentSkill review signals, maintenance freshness, and install readiness. They are a shortlist signal, not a replacement for human review.
Quality
StrongSolid option that is likely worth shortlisting for production workflows.
Trust
Sandbox onlyUseful candidate with missing or mixed trust signals. Keep it in an isolated workspace until the outcome loop proves task fit.
Audit
Needs reviewA machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
OpenAgentSkill Trust Score v5
Run only in a sandbox and compare close alternatives before using it for real work.
Stars
497 GitHub stars
Repo activity
497 stars, 96 forks
Maintenance
1d since push
License
MIT
Install
npx skills add zenstory-ai/video-recap-skills --skill video-script
Install safety
Agent-readable metadata
Use this block or the embedded JSON to decide whether an agent should install this skill, choose an alternative, or ask for human review first.
Suited tasks
Suited agents
Install decision
Trust and risk
Outcome loop
Install command
npx skills add zenstory-ai/video-recap-skills --skill video-scriptDo not use when
Alternative
174.6K Stars
npx skills add anthropics/skills --skill frontend-design
Alternative
84.6K Stars
npx skills add Leonxlnx/taste-skill --skill design-taste-frontend
Alternative
1.8K Stars
npx skills add Alisa0808/vox-director --skill vox-director
Alternative
174.6K Stars
npx skills add anthropics/skills --skill canvas-design
Agent safety v2
Sparse or mixed signals. Useful for discovery, but not for autonomous installation.
Test manually in an isolated workspace and compare against safer alternatives.
high
Skill metadata references terminal, CLI, shell, subprocess, or command execution workflows.
medium
Skill likely fetches remote pages, APIs, repositories, or external services.
medium
Skill may read or write project files, documents, generated artifacts, or local workspace state.
Agent resolve plan
The Resolve API returns the selected skill, alternatives, safety policy, audit notes, install target, and copy-paste prompt an agent can follow without scraping this page.
Open JSON
/api/agent/resolve?task=Use%20video-script%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Resolve text
/api/agent/resolve?task=Use%20video-script%20for%20an%20agent%20workflow&agent=codex&max_risk=medium&format=text
Install handoff
/api/skills/zenstory-ai-video-script/install
Agent should check
Copy prompt
Task: Use video-script in this workspace.
Resolve first: https://www.openagentskill.com/api/agent/resolve?task=Use%20video-script%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Review install handoff: https://www.openagentskill.com/api/skills/zenstory-ai-video-script/install
Install command: npx skills add zenstory-ai/video-recap-skills --skill video-script
Before running it, summarize audit warnings, required permissions, and the fallback skill if install is risky.Agent handoff
Use the public install endpoint to fetch the command, safety checklist, target prompts, and canonical links for this skill.
Install handoff
/api/skills/zenstory-ai-video-script/install
LLM text format
/api/skills/zenstory-ai-video-script/install?format=text
Find alternatives
/api/skills/search?q=video-script&limit=3
Agent prompt
Use video-script for this task. Review https://www.openagentskill.com/api/skills/zenstory-ai-video-script/install, then install with: npx skills add zenstory-ai/video-recap-skills --skill video-scriptRegistry metadata
This page exposes the same decision, trust, audit, use-case, and install signals through the Registry API, so agents can rank this skill without scraping the UI.
Manifest
/api/registry/manifest/zenstory-ai-video-script
LLM text
/api/registry/manifest/zenstory-ai-video-script?format=text
Install alias
/api/registry/install/zenstory-ai-video-script
Recommend
/api/registry/recommend?task=Use%20video-script%20in%20an%20agent%20workflow&limit=3
Agent fit
Research agents
Use-case tags
Platforms
Claude Code
Audit report
A machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
Agent decision cockpit
Shortlist this skill and compare it with close alternatives before production adoption.
Role in stack
Companion skill
Primary fit
Research agents
Trust label
Strong shortlist
Install path
Command ready
Use when
Evidence
review first
Implementation path
Trust profile
Useful candidate with missing or mixed trust signals. Keep it in an isolated workspace until the outcome loop proves task fit.
GitHub adoption
INFO497 GitHub stars
Stars/forks activity
INFO497 stars, 96 forks; issue activity unavailable in current metadata
Recent maintenance
PASS1d since push
License clarity
PASSMIT
Good signals
Review before install
Recommended action
Run only in a sandbox and compare close alternatives before using it for real work.
Quality profile
Solid option that is likely worth shortlisting for production workflows.
Workflow fit
Investigate faster
I need my agent to research a topic, compare sources, and produce a concise report.
Process rich media
I need my agent to process images, video, or audio and extract useful information.
Search private knowledge
I need my agent to build a RAG workflow over documents and retrieve reliable context.
Workflow fit
Find, compare, and synthesize
A workflow for agents that gather sources, compare claims, summarize long material, and draft useful research briefs.
Ingest, retrieve, and cite
A workflow for document-heavy agents that ingest files, create searchable knowledge, retrieve relevant context, and answer with grounded sources.
Design, build, test, and ship interfaces
A practical workflow for agents that turn product briefs or Figma designs into polished frontend code, review the result, test it in a browser, and prepare a safe deployment.
Alternative shortlist
Similar skills that may fit this task.
Guidance for distinctive, intentional UI design, typography, visual direction, and non-template-like product interfaces.
Design and implementation guidance for distinctive landing pages, portfolios, product demos, and purposeful redesigns.
Turn one topic into a narrated Vox-style paper-collage explainer or ad video, from script through captions.
Create original visual art, posters, PNG assets, and PDF documents through a clear design philosophy.
--- name: video-script description: > 对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时使用。适用于故事方向、片段选择、画面/原声/旁白分工、 解说写作与复核。输入 work_dir 中的理解索引;输出 recap_story_plan.json、visual_audio_board.json、 可选 style_card.json、cut 模式需要的 clip_plan.json,以及通过校验的 narration.json。触发词:解说词、写解说、视频旁白、 narration script、写稿、解说文案、剪辑思路、导演思路。 ---
## 1. 定位
本技能负责:创作方向、画面/声音计划、旁白写作与校验。Agent 不是 JSON 填写器,而要依次扮演:
1. 导演 2. 故事编辑 3. 画面剪辑师 4. 声音/旁白编辑 5. 第一次观看的观众
Agent 先记录简洁决定,再写时间线产物。`validate.py` 负责对理解索引做机械校验;full 模式还会把旁白对齐到安静窗口。
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。本技能不从其他技能目录读取参考文件或辅助脚本;外部输入只来自显式路径与 `work_dir` 产物。
### 1.1 创作控制模式
先根据用户要求和 `work_dir` 判断本轮模式;它不是 `full|cut|dub` 渲染模式:
- **CREATE**:首次创作。比较至少两个真正可行的故事/剪辑假设后再选择。 - **DIRECTED**:用户已指定结构、镜头、台词或表达。忠实落实,不为满足“创作流程”虚构替代方案。 - **REVISION**:用户针对已有版本看片修改。最新反馈是当前事实来源;未点名部分默认冻结。
REVISION 先明确本轮修改项与冻结项,再编辑对应层:表达、口语节奏、字幕反馈更新 `style_card.json`;镜头、入出点、表演和声音分工更新 `visual_audio_board.json`;只有观众承诺、POV、主线或 beat 改变时才更新 `recap_story_plan.json`。被删除的镜头、原声或文案也要从相关计划中删除,不能保留过期锚点。不要把看片修改重新做成一次 CREATE。
## 2. 读取素材并确认状态
首先阅读:
- `work_dir/agent_narration_brief.md`:场景、时长、安静窗口与字数预算。 - `asr_writing_chunks.json`:长对白的写作分块。 - `timeline_fusion.json`:判断某段是否有对白或静音槽。 - `vlm_analysis.json` / `asr_result.json`:核对具体画面与原声证据。 - brief 顶部列出的 contact sheet:不要只依赖场景摘要;反应、走位、静止和台词前后的具体时刻常常更重要。
full 模式使用原片时间。cut 模式第一阶段只写 `clip_plan.json`;`edited_source.mp4` 产生后,第二阶段才按输出时间写 `narration.json`。
写任何创作产物前,直接读取 `work_dir` 判断当前阶段:
- `recap_run_manifest.json`:确认 `edit_mode`、源视频和本轮设置。 - full 模式:没有 `narration.json` 时进入写稿;存在时先复核再校验。 - cut 第一阶段:尚无 `clip_plan_validated.json` / `edited_source.mp4`,只写 `clip_plan.json`。 - cut 第二阶段:两者都存在,按 `clip_plan_validated.json.clips[]` 中的 `source_start/end` 与 `output_start/end` 核对映射,再写输出时间的旁白。
必须确认旁白没有跨越错误剪辑边界,也没有落进已删除区间。整个判断只依赖 `work_dir` 产物。
## 3. 制定创作方案
先阅读 `references/creative-editing-playbook.md`,再按创作控制模式写或更新工作产物:
1. **`recap_story_plan.json`**:导演意图、CREATE 中至少两个剪辑假设、选定的 POV / 主线,以及由“变化”定义的 beats。DIRECTED / REVISION 不强行新增假设。 2. **`visual_audio_board.json`**:每拍的画面任务、具体表演/反应、入点/出点、`audio_owner`、原声锚点与 `narration_job`。 3. **`style_card.json`(适用时)**:用户当前认可的声音、口语节奏、字幕阅读姿态和明确禁忌。收到表达或字幕反馈后更新原文件,而不是只改最终文案。
只记录决定、证据锚点、被放弃的备选方案和简短理由,不写冗长思维过程。
### 3.1 导演判断
锁定:
- 观众承诺 - POV - 戏剧问题 - 起始与结束情绪 - 隐瞒与揭示 - 结尾余味
### 3.2 故事编辑
CREATE 比较两个真正可行的结构后选择一个;DIRECTED / REVISION 沿用用户指定或已确认的结构,除非最新反馈明确改变故事方向。每个 beat 至少改变一项:知识、权力、目标、关系、情绪或风险。若删除后因果、人物和情绪都没有损失,该 beat 通常不应保留。
### 3.3 画面剪辑
选择具体时刻,而不是只选择事件。比较:
- 说话者与倾听者 - 动作与反应 - 早进与晚进 - 早出与多停半秒
在不破坏理解的前提下晚进早出,同时保留不可替代的表演、停顿、失误、动作声和完整台词。
### 3.4 声音与旁白分工
先指定 `audio_owner`,再写字。旁白只允许承担以下 `narration_job`:
- `context` - `causal_link` - `foreshadow` - `interpretation` - `transition` - `none`
画面、原声或沉默已经足够时使用 `none`,不要默认铺旁白。
### 3.5 cut 模式第一阶段
cut 模式先根据 `recap_story_plan.json` 与 `visual_audio_board.json` 写原片时间的 `clip_plan.json`,此时不要写 `narration.json`:
```json { "target_duration": "10m", "clips": [ { "start": 12.0, "end": 38.0, "reason": "b01 | hook | knowledge: unknown→threat | POV=主角 | 保留倾听反应 | 入点=问题已问出 | 出点=沉默落地" } ] } ```
`reason` 统一使用:
```text beat_id | function | change | POV | preferred moment | 入点 | 出点 ```
片段顺序必须构成一条完整故事线,而不是无序高光。可使用 0–1 个 cold open,随后回到因果清楚的 setup → turn → escalation → payoff。片段长度服从具体时刻,不使用统一秒数模板;片尾必须保留完整台词或动作。对短时间内密集的 scene-change 候选,先区分原片切点与本次拼接点:原片无关短镜头整段删,相关短镜头扩展到完整动作/反应;本次拼接点优先移动边界、恢复同源连续运动或合并片段,尽量不制造人工闪切。
## 4. 撰写旁白
full 模式直接按原片时间写;cut 第二阶段先查看 `edited_source.mp4` 与剪后故事板,补充 `visual_audio_board.json` 的输出时间并重新确认 `audio_owner` / `narration_job`,再按输出时间写:
```json [ { "start": 5.0, "end": 12.0, "narration": "解说文本。", "pause_after_ms": 250, "overlaps_speech": true, "emotion": "紧张" } ] ```
字段说明:
| 字段 | 含义 | |------|------| | `start` / `end` | full 模式为原片时间;cut 第二阶段为输出时间 | | `narration` | 解说文本 | | `pause_after_ms` | 段后停顿,默认 250ms | | `overlaps_speech` | 是否与原对白重叠;连续铺底窗口通常为 `true`,真正静音槽才为 `false` | | `emotion` | 整个解说块的 MiMo TTS 情绪/语气标签 |
### 4.1 写作规则
1. **先有 `narration_job`,后有句子**:没有明确任务就不写;旁白不是默认音轨。 2. **按连续思路写**:旁白拥有一个 beat 时,用一个或少量完整句子完成“前提 → 触发动作 → 变化/意义”,并在一次 TTS 中合成。句号服从口语思路和呼吸,不服从字幕换行;不要固定句数,也不要“一句一停”。 3. **7:3 不是配额**:只在素材判断不足时作为避免墙到墙旁白的粗略首稿参考。实际比例服从 `audio_owner`;强对白、动作声或沉默可以完整拥有一个 beat。 4. **视听接力**:旁白若引出原声,块尾要让观众想听;原声结束后的下一块要承接它造成的变化。 5. **按有效语速控量**:用 `字数 / brief 头部 speech budget` 估算窗口;装不下时删减或拆分叙事任务,不用加速堆字。 6. **不看图说话**:旁白只增加上下文、因果、预期、证据支持的解释或跨越。 7. **人物与证据优先**:优先使用已知角色名;关系、动机、潜台词和结果必须指向 visual / ASR / research / user context,且不能把背景资料伪装成当前画面事实。 8. **写给耳朵听**:使用具体名词和动词,句子完整、口语可听;避免字幕腔、半句、空泛拔高和破折号。TTS 文本先保证听感连续,字幕再按阅读宽度拆分,不能反过来把朗读稿切碎。 9. **避免模板化纠偏**:“不是 A,而是 B”只在确实存在一个观众可能相信、而素材又要纠正的判断时使用。它不是禁句,但不能靠先否定再肯定制造假洞察;优先直接写人物的动作、因果和后果。
### 4.2 解说结构
- **钩子**:提出正文会真实兑现的问题或利害,不用无关留存话术。 - **主线**:围绕选定 POV 与主线推进,不在每个场景重新开篇。 - **递进**:后续 beat 必须提高风险、改变关系或提供新信息。 - **悬念缺口**:只预告之后真的会回收的后果。 - **收尾**:回答或有意转化开头问题,留下明确余味。 - **衔接**:旁白块与相邻原声属于同一个 beat,前者铺垫、后者呈现、下一块承接。
### 4.3 原声留白字幕
可选写 `original_subtitles.json`,使用成片输出时间:
```json [{"start": 15.0, "end": 17.0, "text": "原声台词"}] ```
只写留白中实际听得到的台词,订正 ASR 错字与人名,每条尽量控制在一行;被旁白盖住或已经剪掉的句子不要写。省略时,合成阶段会使用保守的 ASR 映射兜底,并在成片中用 `「」` 区分原声对白与旁白。
## 5. 创作自审
在调用 LLM 评审前做以下**反事实检查**:
1. 删除每个 beat:若因果、人物、情绪或承诺没有损失,就删除。 2. 比较说话者/动作与倾听者/反应:保留更符合 POV 和情绪的时刻。 3. 静音旁白:画面与原声仍应承载可见行动、人物行为和关键情绪。 4. 只听声音:旁白应形成可听懂的主线,而不是画面字幕。 5. 把旁白换成原声或沉默:若场景自身更有力量,就让出声音所有权。 6. 检查开头问题是否真实,结尾是否回答或转化它。 7. 检查相邻短句能否合成一个连续思路;字幕分行不得成为 TTS 断句理由。
只记录并优先修复 1–3 个回报最高的问题;先改结构,再润色句子。
REVISION 还要逐项确认:用户点名的问题已经改变,未点名的冻结项没有意外变化,相关 `style_card.json` / `visual_audio_board.json` 中不存在旧镜头或旧表达。除非用户要求备选版本,不额外扩展新方向。
## 6. 评审与校验
### 6.1 建议型语义评审
```bash python3 scripts/review.py --work-dir <work_dir> ```
评审会自动识别 cut 模式,并在存在已校验剪辑计划时按输出时间线核对;`--timeline source` 可强制使用原片时间。打开 `narration_review.md`,逐项处理 `error`,尤其是 `category=hallucination`。
重复修改并评审,直到:
- `verdict` 为 `PASS` / `OK` 且没有 `error`;或 - 对仍保留的问题做明确 override。
覆盖决定追加到 `work_dir/narration_review_override.md`:
```markdown ### 覆盖记录 — <date> - 问题:segment 4 / category=hallucination - 评审意见:“他早已知情”缺少画面/对白依据 - 决定:KEEP — 该事实来自用户提供的当前集背景,而非未来剧情 - 签署:<agent/human> ```
`review.py` 本身只写报告,默认调用策略为建议型、失败开放;若调用方显式开启严格评审,事实矛盾、残句、解析失败或评审不可用可在 TTS 前阻断。覆盖记录只用于审计,`review.py` / `validate.py` 不读取它。
### 6.2 确定性硬校验
```bash python3 scripts/validate.py --work-dir <work_dir> --mode full # cut 输出时间线由编排器使用 --mode cut_output ```
命令写出 `narration_lint.json`。full 模式还会根据安静窗口重写 `narration.json` 的时间。修复所有 error 后重复运行,直到校验干净,再继续 TTS 与合成。
片名或题材明确但缺少剧情上下文时,先按本技能的 `references/research-guide.md` 写 `background_research.json`。若理解素材偏薄,brief 中的数量只能当上限:宁可少写、写实,也不要为凑数复述画面。
## 7. 能力边界
- 不运行 ASR / VLM;只消费视频理解索引。 - 不合成 TTS,也不渲染视频。 - 不根据平台分析做优化;先建立内容意图与剪辑一致性。 - `review.py` 不改写 `narration.json`;是否采用严格门禁由调用方决定。 - `validate.py` 不改写文本含义,只检查或对齐时间与安静窗口。
Source provenance
Decision snapshot
recent repository activity
Audit
Install and adoption review
Agent-proven evidence
Outcome reports after resolve, review, install, and one narrow run.
No agent outcome data yet. The first agent run can report success, setup needs, risk blocks, failure, or not-relevant through /api/agent/outcome.
Install
Free and open source. Review the report before installing into production agents.
Growth loop
Scenario-led draft for video-script, ready for a manual X post.
video-script: 对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时... 497 stars https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=x
Listing + install path for video-script: https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=x Install: npx skills add zenstory-ai/video-recap-skills --skill video-script
Listing source
This listing was indexed from public sources and is not marked official until a maintainer claim is approved.
Attribution links to the public repository or creator profile. Creators can claim the listing to update ownership signals.
Claim this skillOwner claim
This Registry indexed listing is attributed to zenstory-ai but is not marked official yet. Claim it to add a verified owner signal and make future launch, install, and audit updates easier to trust.
Creator backlink kit
Show the canonical listing, current trust and audit signals, and real Agent-Proven evidence where developers evaluate the repository.
[](https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/zenstory-ai-video-script/audit)
[](https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)zenstory-ai
@zenstory-ai
Share whether this skill looks useful for your agent workflow. Aggregated feedback improves rankings over time.
Sandbox only
Frontend Design
Guidance for distinctive, intentional UI design, typography, visual direction, and non-template-like product interfaces.
174.6K StarsTaste Skill: Anti-Slop Frontend
Design and implementation guidance for distinctive landing pages, portfolios, product demos, and purposeful redesigns.
84.6K StarsVox Director
Turn one topic into a narrated Vox-style paper-collage explainer or ad video, from script through captions.
1.8K StarsCanvas Design
Create original visual art, posters, PNG assets, and PDF documents through a clear design philosophy.
174.6K StarsSandbox only
Install targets
Codex install prompt
Install the "video-script" agent skill from https://github.com/zenstory-ai/video-recap-skills/tree/main/skills/video-script. Read its SKILL.md or equivalent instructions first, install only the files needed for this workspace, and summarize any required setup before using it. Skill purpose: 对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时使用。适用于故事方向、片段选择、画面/原声/旁白分工、 解说写作与复核。输入 work_dir 中的理解索引;输出 recap_story_plan.json、visual_audio_board.json、 可选 style_card.json、cut 模式需要的 clip_plan.json,以及通过校验的 narration.json。触发词:解说词、写解说、视频旁白、 narration script、写稿、解说文案、剪辑思路、导演思路。 After the install attempt, report the result to https://www.openagentskill.com/api/agent/outcome with POST JSON {"event_id":"install_<unique-id>","skill_slug":"zenstory-ai-video-script","task":"Install video-script","agent":"codex","outcome":"success","install_used":true}. Replace event_id with a unique value and outcome with success or failed. Report success only after the skill is installed and a minimal verification passes.Supply asset profile
Design assets, images, video, audio, multimodal media, presentation, and creative production skills.
Scenario
Multimodal media
I need my agent to process images, video, or audio and extract useful information.
Agent fit
Claude Code + CLI + Codex
Codex, Claude Code, Cursor, CLI, or custom agents.
Install
Ready
npx skills add zenstory-ai/video-recap-skills --skill video-script
Maintenance
fresh
1d since push
Risk
Needs review
The skill depends on several Python scripts and local modules, but SKILL.md does not document installation steps, required dependencies, or the exact validate.py invocation command.
GitHub quality
497
74/100 Quality · 71/100 Trust
Coverage tags
Review notes
The skill depends on several Python scripts and local modules, but SKILL.md does not document installation steps, required dependencies, or the exact validate.py invocation command. · It reads untrusted JSON artifacts from work_dir and may perform optional web research; there is no explicit note that work_dir evidence files must be treated as data, not as instructions.
Agent adoption scorecard
These scores combine public repository metadata, OpenAgentSkill review signals, maintenance freshness, and install readiness. They are a shortlist signal, not a replacement for human review.
Quality
StrongSolid option that is likely worth shortlisting for production workflows.
Trust
Sandbox onlyUseful candidate with missing or mixed trust signals. Keep it in an isolated workspace until the outcome loop proves task fit.
Audit
Needs reviewA machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
OpenAgentSkill Trust Score v5
Run only in a sandbox and compare close alternatives before using it for real work.
Stars
497 GitHub stars
Repo activity
497 stars, 96 forks
Maintenance
1d since push
License
MIT
Install
npx skills add zenstory-ai/video-recap-skills --skill video-script
Install safety
Agent-readable metadata
Use this block or the embedded JSON to decide whether an agent should install this skill, choose an alternative, or ask for human review first.
Suited tasks
Suited agents
Install decision
Trust and risk
Outcome loop
Install command
npx skills add zenstory-ai/video-recap-skills --skill video-scriptDo not use when
Alternative
174.6K Stars
npx skills add anthropics/skills --skill frontend-design
Alternative
84.6K Stars
npx skills add Leonxlnx/taste-skill --skill design-taste-frontend
Alternative
1.8K Stars
npx skills add Alisa0808/vox-director --skill vox-director
Alternative
174.6K Stars
npx skills add anthropics/skills --skill canvas-design
Agent safety v2
Sparse or mixed signals. Useful for discovery, but not for autonomous installation.
Test manually in an isolated workspace and compare against safer alternatives.
high
Skill metadata references terminal, CLI, shell, subprocess, or command execution workflows.
medium
Skill likely fetches remote pages, APIs, repositories, or external services.
medium
Skill may read or write project files, documents, generated artifacts, or local workspace state.
Agent resolve plan
The Resolve API returns the selected skill, alternatives, safety policy, audit notes, install target, and copy-paste prompt an agent can follow without scraping this page.
Open JSON
/api/agent/resolve?task=Use%20video-script%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Resolve text
/api/agent/resolve?task=Use%20video-script%20for%20an%20agent%20workflow&agent=codex&max_risk=medium&format=text
Install handoff
/api/skills/zenstory-ai-video-script/install
Agent should check
Copy prompt
Task: Use video-script in this workspace.
Resolve first: https://www.openagentskill.com/api/agent/resolve?task=Use%20video-script%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Review install handoff: https://www.openagentskill.com/api/skills/zenstory-ai-video-script/install
Install command: npx skills add zenstory-ai/video-recap-skills --skill video-script
Before running it, summarize audit warnings, required permissions, and the fallback skill if install is risky.Agent handoff
Use the public install endpoint to fetch the command, safety checklist, target prompts, and canonical links for this skill.
Install handoff
/api/skills/zenstory-ai-video-script/install
LLM text format
/api/skills/zenstory-ai-video-script/install?format=text
Find alternatives
/api/skills/search?q=video-script&limit=3
Agent prompt
Use video-script for this task. Review https://www.openagentskill.com/api/skills/zenstory-ai-video-script/install, then install with: npx skills add zenstory-ai/video-recap-skills --skill video-scriptRegistry metadata
This page exposes the same decision, trust, audit, use-case, and install signals through the Registry API, so agents can rank this skill without scraping the UI.
Manifest
/api/registry/manifest/zenstory-ai-video-script
LLM text
/api/registry/manifest/zenstory-ai-video-script?format=text
Install alias
/api/registry/install/zenstory-ai-video-script
Recommend
/api/registry/recommend?task=Use%20video-script%20in%20an%20agent%20workflow&limit=3
Agent fit
Research agents
Use-case tags
Platforms
Claude Code
Audit report
A machine-readable review of install readiness, security metadata, maintenance, and adoption risk.
Agent decision cockpit
Shortlist this skill and compare it with close alternatives before production adoption.
Role in stack
Companion skill
Primary fit
Research agents
Trust label
Strong shortlist
Install path
Command ready
Use when
Evidence
review first
Implementation path
Trust profile
Useful candidate with missing or mixed trust signals. Keep it in an isolated workspace until the outcome loop proves task fit.
GitHub adoption
INFO497 GitHub stars
Stars/forks activity
INFO497 stars, 96 forks; issue activity unavailable in current metadata
Recent maintenance
PASS1d since push
License clarity
PASSMIT
Good signals
Review before install
Recommended action
Run only in a sandbox and compare close alternatives before using it for real work.
Quality profile
Solid option that is likely worth shortlisting for production workflows.
Workflow fit
Investigate faster
I need my agent to research a topic, compare sources, and produce a concise report.
Process rich media
I need my agent to process images, video, or audio and extract useful information.
Search private knowledge
I need my agent to build a RAG workflow over documents and retrieve reliable context.
Workflow fit
Find, compare, and synthesize
A workflow for agents that gather sources, compare claims, summarize long material, and draft useful research briefs.
Ingest, retrieve, and cite
A workflow for document-heavy agents that ingest files, create searchable knowledge, retrieve relevant context, and answer with grounded sources.
Design, build, test, and ship interfaces
A practical workflow for agents that turn product briefs or Figma designs into polished frontend code, review the result, test it in a browser, and prepare a safe deployment.
Alternative shortlist
Similar skills that may fit this task.
Guidance for distinctive, intentional UI design, typography, visual direction, and non-template-like product interfaces.
Design and implementation guidance for distinctive landing pages, portfolios, product demos, and purposeful redesigns.
Turn one topic into a narrated Vox-style paper-collage explainer or ad video, from script through captions.
Create original visual art, posters, PNG assets, and PDF documents through a clear design philosophy.
--- name: video-script description: > 对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时使用。适用于故事方向、片段选择、画面/原声/旁白分工、 解说写作与复核。输入 work_dir 中的理解索引;输出 recap_story_plan.json、visual_audio_board.json、 可选 style_card.json、cut 模式需要的 clip_plan.json,以及通过校验的 narration.json。触发词:解说词、写解说、视频旁白、 narration script、写稿、解说文案、剪辑思路、导演思路。 ---
## 1. 定位
本技能负责:创作方向、画面/声音计划、旁白写作与校验。Agent 不是 JSON 填写器,而要依次扮演:
1. 导演 2. 故事编辑 3. 画面剪辑师 4. 声音/旁白编辑 5. 第一次观看的观众
Agent 先记录简洁决定,再写时间线产物。`validate.py` 负责对理解索引做机械校验;full 模式还会把旁白对齐到安静窗口。
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。本技能不从其他技能目录读取参考文件或辅助脚本;外部输入只来自显式路径与 `work_dir` 产物。
### 1.1 创作控制模式
先根据用户要求和 `work_dir` 判断本轮模式;它不是 `full|cut|dub` 渲染模式:
- **CREATE**:首次创作。比较至少两个真正可行的故事/剪辑假设后再选择。 - **DIRECTED**:用户已指定结构、镜头、台词或表达。忠实落实,不为满足“创作流程”虚构替代方案。 - **REVISION**:用户针对已有版本看片修改。最新反馈是当前事实来源;未点名部分默认冻结。
REVISION 先明确本轮修改项与冻结项,再编辑对应层:表达、口语节奏、字幕反馈更新 `style_card.json`;镜头、入出点、表演和声音分工更新 `visual_audio_board.json`;只有观众承诺、POV、主线或 beat 改变时才更新 `recap_story_plan.json`。被删除的镜头、原声或文案也要从相关计划中删除,不能保留过期锚点。不要把看片修改重新做成一次 CREATE。
## 2. 读取素材并确认状态
首先阅读:
- `work_dir/agent_narration_brief.md`:场景、时长、安静窗口与字数预算。 - `asr_writing_chunks.json`:长对白的写作分块。 - `timeline_fusion.json`:判断某段是否有对白或静音槽。 - `vlm_analysis.json` / `asr_result.json`:核对具体画面与原声证据。 - brief 顶部列出的 contact sheet:不要只依赖场景摘要;反应、走位、静止和台词前后的具体时刻常常更重要。
full 模式使用原片时间。cut 模式第一阶段只写 `clip_plan.json`;`edited_source.mp4` 产生后,第二阶段才按输出时间写 `narration.json`。
写任何创作产物前,直接读取 `work_dir` 判断当前阶段:
- `recap_run_manifest.json`:确认 `edit_mode`、源视频和本轮设置。 - full 模式:没有 `narration.json` 时进入写稿;存在时先复核再校验。 - cut 第一阶段:尚无 `clip_plan_validated.json` / `edited_source.mp4`,只写 `clip_plan.json`。 - cut 第二阶段:两者都存在,按 `clip_plan_validated.json.clips[]` 中的 `source_start/end` 与 `output_start/end` 核对映射,再写输出时间的旁白。
必须确认旁白没有跨越错误剪辑边界,也没有落进已删除区间。整个判断只依赖 `work_dir` 产物。
## 3. 制定创作方案
先阅读 `references/creative-editing-playbook.md`,再按创作控制模式写或更新工作产物:
1. **`recap_story_plan.json`**:导演意图、CREATE 中至少两个剪辑假设、选定的 POV / 主线,以及由“变化”定义的 beats。DIRECTED / REVISION 不强行新增假设。 2. **`visual_audio_board.json`**:每拍的画面任务、具体表演/反应、入点/出点、`audio_owner`、原声锚点与 `narration_job`。 3. **`style_card.json`(适用时)**:用户当前认可的声音、口语节奏、字幕阅读姿态和明确禁忌。收到表达或字幕反馈后更新原文件,而不是只改最终文案。
只记录决定、证据锚点、被放弃的备选方案和简短理由,不写冗长思维过程。
### 3.1 导演判断
锁定:
- 观众承诺 - POV - 戏剧问题 - 起始与结束情绪 - 隐瞒与揭示 - 结尾余味
### 3.2 故事编辑
CREATE 比较两个真正可行的结构后选择一个;DIRECTED / REVISION 沿用用户指定或已确认的结构,除非最新反馈明确改变故事方向。每个 beat 至少改变一项:知识、权力、目标、关系、情绪或风险。若删除后因果、人物和情绪都没有损失,该 beat 通常不应保留。
### 3.3 画面剪辑
选择具体时刻,而不是只选择事件。比较:
- 说话者与倾听者 - 动作与反应 - 早进与晚进 - 早出与多停半秒
在不破坏理解的前提下晚进早出,同时保留不可替代的表演、停顿、失误、动作声和完整台词。
### 3.4 声音与旁白分工
先指定 `audio_owner`,再写字。旁白只允许承担以下 `narration_job`:
- `context` - `causal_link` - `foreshadow` - `interpretation` - `transition` - `none`
画面、原声或沉默已经足够时使用 `none`,不要默认铺旁白。
### 3.5 cut 模式第一阶段
cut 模式先根据 `recap_story_plan.json` 与 `visual_audio_board.json` 写原片时间的 `clip_plan.json`,此时不要写 `narration.json`:
```json { "target_duration": "10m", "clips": [ { "start": 12.0, "end": 38.0, "reason": "b01 | hook | knowledge: unknown→threat | POV=主角 | 保留倾听反应 | 入点=问题已问出 | 出点=沉默落地" } ] } ```
`reason` 统一使用:
```text beat_id | function | change | POV | preferred moment | 入点 | 出点 ```
片段顺序必须构成一条完整故事线,而不是无序高光。可使用 0–1 个 cold open,随后回到因果清楚的 setup → turn → escalation → payoff。片段长度服从具体时刻,不使用统一秒数模板;片尾必须保留完整台词或动作。对短时间内密集的 scene-change 候选,先区分原片切点与本次拼接点:原片无关短镜头整段删,相关短镜头扩展到完整动作/反应;本次拼接点优先移动边界、恢复同源连续运动或合并片段,尽量不制造人工闪切。
## 4. 撰写旁白
full 模式直接按原片时间写;cut 第二阶段先查看 `edited_source.mp4` 与剪后故事板,补充 `visual_audio_board.json` 的输出时间并重新确认 `audio_owner` / `narration_job`,再按输出时间写:
```json [ { "start": 5.0, "end": 12.0, "narration": "解说文本。", "pause_after_ms": 250, "overlaps_speech": true, "emotion": "紧张" } ] ```
字段说明:
| 字段 | 含义 | |------|------| | `start` / `end` | full 模式为原片时间;cut 第二阶段为输出时间 | | `narration` | 解说文本 | | `pause_after_ms` | 段后停顿,默认 250ms | | `overlaps_speech` | 是否与原对白重叠;连续铺底窗口通常为 `true`,真正静音槽才为 `false` | | `emotion` | 整个解说块的 MiMo TTS 情绪/语气标签 |
### 4.1 写作规则
1. **先有 `narration_job`,后有句子**:没有明确任务就不写;旁白不是默认音轨。 2. **按连续思路写**:旁白拥有一个 beat 时,用一个或少量完整句子完成“前提 → 触发动作 → 变化/意义”,并在一次 TTS 中合成。句号服从口语思路和呼吸,不服从字幕换行;不要固定句数,也不要“一句一停”。 3. **7:3 不是配额**:只在素材判断不足时作为避免墙到墙旁白的粗略首稿参考。实际比例服从 `audio_owner`;强对白、动作声或沉默可以完整拥有一个 beat。 4. **视听接力**:旁白若引出原声,块尾要让观众想听;原声结束后的下一块要承接它造成的变化。 5. **按有效语速控量**:用 `字数 / brief 头部 speech budget` 估算窗口;装不下时删减或拆分叙事任务,不用加速堆字。 6. **不看图说话**:旁白只增加上下文、因果、预期、证据支持的解释或跨越。 7. **人物与证据优先**:优先使用已知角色名;关系、动机、潜台词和结果必须指向 visual / ASR / research / user context,且不能把背景资料伪装成当前画面事实。 8. **写给耳朵听**:使用具体名词和动词,句子完整、口语可听;避免字幕腔、半句、空泛拔高和破折号。TTS 文本先保证听感连续,字幕再按阅读宽度拆分,不能反过来把朗读稿切碎。 9. **避免模板化纠偏**:“不是 A,而是 B”只在确实存在一个观众可能相信、而素材又要纠正的判断时使用。它不是禁句,但不能靠先否定再肯定制造假洞察;优先直接写人物的动作、因果和后果。
### 4.2 解说结构
- **钩子**:提出正文会真实兑现的问题或利害,不用无关留存话术。 - **主线**:围绕选定 POV 与主线推进,不在每个场景重新开篇。 - **递进**:后续 beat 必须提高风险、改变关系或提供新信息。 - **悬念缺口**:只预告之后真的会回收的后果。 - **收尾**:回答或有意转化开头问题,留下明确余味。 - **衔接**:旁白块与相邻原声属于同一个 beat,前者铺垫、后者呈现、下一块承接。
### 4.3 原声留白字幕
可选写 `original_subtitles.json`,使用成片输出时间:
```json [{"start": 15.0, "end": 17.0, "text": "原声台词"}] ```
只写留白中实际听得到的台词,订正 ASR 错字与人名,每条尽量控制在一行;被旁白盖住或已经剪掉的句子不要写。省略时,合成阶段会使用保守的 ASR 映射兜底,并在成片中用 `「」` 区分原声对白与旁白。
## 5. 创作自审
在调用 LLM 评审前做以下**反事实检查**:
1. 删除每个 beat:若因果、人物、情绪或承诺没有损失,就删除。 2. 比较说话者/动作与倾听者/反应:保留更符合 POV 和情绪的时刻。 3. 静音旁白:画面与原声仍应承载可见行动、人物行为和关键情绪。 4. 只听声音:旁白应形成可听懂的主线,而不是画面字幕。 5. 把旁白换成原声或沉默:若场景自身更有力量,就让出声音所有权。 6. 检查开头问题是否真实,结尾是否回答或转化它。 7. 检查相邻短句能否合成一个连续思路;字幕分行不得成为 TTS 断句理由。
只记录并优先修复 1–3 个回报最高的问题;先改结构,再润色句子。
REVISION 还要逐项确认:用户点名的问题已经改变,未点名的冻结项没有意外变化,相关 `style_card.json` / `visual_audio_board.json` 中不存在旧镜头或旧表达。除非用户要求备选版本,不额外扩展新方向。
## 6. 评审与校验
### 6.1 建议型语义评审
```bash python3 scripts/review.py --work-dir <work_dir> ```
评审会自动识别 cut 模式,并在存在已校验剪辑计划时按输出时间线核对;`--timeline source` 可强制使用原片时间。打开 `narration_review.md`,逐项处理 `error`,尤其是 `category=hallucination`。
重复修改并评审,直到:
- `verdict` 为 `PASS` / `OK` 且没有 `error`;或 - 对仍保留的问题做明确 override。
覆盖决定追加到 `work_dir/narration_review_override.md`:
```markdown ### 覆盖记录 — <date> - 问题:segment 4 / category=hallucination - 评审意见:“他早已知情”缺少画面/对白依据 - 决定:KEEP — 该事实来自用户提供的当前集背景,而非未来剧情 - 签署:<agent/human> ```
`review.py` 本身只写报告,默认调用策略为建议型、失败开放;若调用方显式开启严格评审,事实矛盾、残句、解析失败或评审不可用可在 TTS 前阻断。覆盖记录只用于审计,`review.py` / `validate.py` 不读取它。
### 6.2 确定性硬校验
```bash python3 scripts/validate.py --work-dir <work_dir> --mode full # cut 输出时间线由编排器使用 --mode cut_output ```
命令写出 `narration_lint.json`。full 模式还会根据安静窗口重写 `narration.json` 的时间。修复所有 error 后重复运行,直到校验干净,再继续 TTS 与合成。
片名或题材明确但缺少剧情上下文时,先按本技能的 `references/research-guide.md` 写 `background_research.json`。若理解素材偏薄,brief 中的数量只能当上限:宁可少写、写实,也不要为凑数复述画面。
## 7. 能力边界
- 不运行 ASR / VLM;只消费视频理解索引。 - 不合成 TTS,也不渲染视频。 - 不根据平台分析做优化;先建立内容意图与剪辑一致性。 - `review.py` 不改写 `narration.json`;是否采用严格门禁由调用方决定。 - `validate.py` 不改写文本含义,只检查或对齐时间与安静窗口。
Source provenance
Decision snapshot
recent repository activity
Audit
Install and adoption review
Agent-proven evidence
Outcome reports after resolve, review, install, and one narrow run.
No agent outcome data yet. The first agent run can report success, setup needs, risk blocks, failure, or not-relevant through /api/agent/outcome.
Install
Free and open source. Review the report before installing into production agents.
Growth loop
Scenario-led draft for video-script, ready for a manual X post.
video-script: 对已完成分析的视频进行导演与剪辑策划,再写带时间戳的中文解说并校验。work_dir 已包含 agent_narration_brief.md 与 vlm_analysis.json 时... 497 stars https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=x
Listing + install path for video-script: https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=x Install: npx skills add zenstory-ai/video-recap-skills --skill video-script
Listing source
This listing was indexed from public sources and is not marked official until a maintainer claim is approved.
Attribution links to the public repository or creator profile. Creators can claim the listing to update ownership signals.
Claim this skillOwner claim
This Registry indexed listing is attributed to zenstory-ai but is not marked official yet. Claim it to add a verified owner signal and make future launch, install, and audit updates easier to trust.
Creator backlink kit
Show the canonical listing, current trust and audit signals, and real Agent-Proven evidence where developers evaluate the repository.
[](https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)
[](https://www.openagentskill.com/skills/zenstory-ai-video-script/audit)
[](https://www.openagentskill.com/skills/zenstory-ai-video-script?ref=github&utm_source=github&utm_medium=referral&utm_campaign=creator_badge)zenstory-ai
@zenstory-ai
Share whether this skill looks useful for your agent workflow. Aggregated feedback improves rankings over time.
Sandbox only
Frontend Design
Guidance for distinctive, intentional UI design, typography, visual direction, and non-template-like product interfaces.
174.6K StarsTaste Skill: Anti-Slop Frontend
Design and implementation guidance for distinctive landing pages, portfolios, product demos, and purposeful redesigns.
84.6K StarsVox Director
Turn one topic into a narrated Vox-style paper-collage explainer or ad video, from script through captions.
1.8K StarsCanvas Design
Create original visual art, posters, PNG assets, and PDF documents through a clear design philosophy.
174.6K StarsPermission surface
shell or command execution
Agent outcomes
No agent outcome data yet
Docs
Usable metadata, review docs
Risk summary
Install readiness
Permission surface
shell or command execution
Agent outcomes
No agent outcome data yet
Docs
Usable metadata, review docs
Risk summary
Install readiness
Permission surface
shell or command execution
Agent outcomes
No agent outcome data yet
Docs
Usable metadata, review docs
Risk summary
Install readiness
Permission surface
shell or command execution
Agent outcomes
No agent outcome data yet
Docs
Usable metadata, review docs
Risk summary
Install readiness