norahe0304-art/30x-video / cli/skill
30x-web-to-video
FlaggedUse when a user gives a brand URL and wants a premium launch, demo, or marketing video. Zero-manual workflow — Claude runs the V2 orchestrator itself (scrapes brand evidence, auto-installs missing tools, generates the first-cut Remotion project). User only sees the URL and the Studio preview.
scripts/AGENTS.md
View on GitHub# scripts/
> L2 | 父级: [../AGENTS.md](../AGENTS.md)
成员清单
beat-sync.ts: BGM 节拍检测与默认切点建议,输出 `beat-map.json`,给 TransitionSeries 对齐节奏。
timing-audit.ts: 扫描场景时长与动画完成时机,输出逐场 timing audit 表,阻断过短场景。
visual-audit.ts: 聚合字体、safe-zone、真实素材、可读性与视觉规则,作为 render 前阻断审计。
benchmark-suite.ts: 读取 canonical benchmark manifest,批量执行 URL intake,并输出 per-benchmark 结果 JSON 与 suite summary。
evidence-model.ts: URL intake V2 的共享证据模型、评分类型与 mode 判定辅助。
scene-constitution.ts: 从 BrandReport 推 archetype、story intent 和双引擎 scene constitution。
project-blueprint.ts: 复制 scaffold 并写入 `brand-report.json`、`scene-constitution.json`、`asset-manifest.json`、`story.md`、`review.md` 与生成数据文件。
url-to-video.ts: One-URL orchestrator;完成 preflight(ffmpeg/yt-dlp/aubiotrack/playwright 交互式安装)、evidence harvest(playwright 优先,agent-browser 兜底)、评分、mode selection、scene constitution、工程输出,并自动跑 beat-sync + visual-audit 写回 review.md。CLI: `--out` `--force` `--offline` `--yes/-y` `--no-install`。
critique-scenes.ts: 品味回路第一环。渲每一幕代表帧 → 喂 `claude -p --json-schema` vision critic → 写 `.iterate/round-N/critique.json`,含 5 维打分和最弱场景的可执行指令。
regenerate-scene.ts: 品味回路第二环。读 critique.json,备份 MainVideo.tsx/theme.ts,起 `claude -p` 只改最弱 Act 组件,跑 `tsc --noEmit`,失败自动回滚。
iterate.ts: 品味回路编排器。循环 critique → regenerate → 下一轮渲帧,终止于 rounds 用尽或 `overallScore >= threshold`,结果落在 `.iterate/round-N/` 与 `.iterate/history.json`。CLI: `iterate <project-dir> [--rounds 3] [--threshold 8]`。
render-qa.ts: 渲染后机器质检(Step 8 阻断门)。ffprobe 流/尺寸/时长 + blackdetect/freezedetect/silencedetect/volumedetect,pass/warn/fail 报告,warn 窗口交给人眼 Read 判定。CLI: `render-qa.ts <mp4> [--expect-duration S] [--json out.json]`。移植自 maxazure/video-editing-skill 的 render_qa.py,阈值按本 skill 重调。
法则: 脚本先验真相·输出必须可阻断·审计语言要具体
[PROTOCOL]: 变更时更新此头部,然后检查 AGENTS.md
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| planning-and-task-breakdownaddyosmani/agent-skillsBreaks work into ordered tasks. Use when you have a spec or clear requirements and need to break work into implementable tasks. Use when a task feels too large to start, when you need to estimate scope, or when parallel work is possible. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| screenpipethedotmack/claude-memSearch screen and audio history recorded by Screenpipe and optionally save selected context into claude-mem. Use when the user asks to recall desktop activity, a meeting, or something seen in another app through Screenpipe. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |
| contact-widgetnexu-io/open-designSelf-contained floating chat widget with welcome screen, social links, meeting button, and message input. Single HTML file, zero dependencies. | Stars: 108k | Updated | Claude CodeCodex | Verified | Install |
| meeting-notesnexu-io/open-designMeeting notes page — title bar with attendees, agenda checklist, decisions block, action items table with owners + dates, and a "next meeting" footer. Use when the brief mentions "meeting notes", "minutes", "1:1 notes", "all-hands recap", or "会议纪要". | Stars: 108k | Updated | Claude CodeCodex | Verified | Install |