feat: add desktop video and ppt skills

This commit is contained in:
boos
2026-06-29 12:54:57 +08:00
parent bdfb9b8dc2
commit be1f388b26
9 changed files with 328 additions and 0 deletions
@@ -0,0 +1,28 @@
---
name: ffmpeg-media
description: FFmpeg media processing workflow for professional video/audio analysis, compression, conversion, clipping, transcoding, timestamp summaries, subtitles, audio extraction, concat, normalization, and command generation. Use when the user asks to inspect, cut, compress, convert, batch process, or repair media files with FFmpeg.
---
# FFmpeg Media Skill
Use this skill when the task needs reliable command-line video or audio processing.
Primary tool: https://ffmpeg.org
Example agent skill: https://github.com/fabriqaai/ffmpeg-analyse-video-skill
## Workflow
1. Inspect the input media first with `ffprobe` when duration, streams, codecs, dimensions, frame rate, or timestamps matter.
2. Preserve quality by copying streams with `-c copy` for simple trims/containers when possible.
3. Re-encode only when required for compression, compatibility, scaling, burn-in subtitles, filters, or codec changes.
4. For social/video platform outputs, specify codec, pixel format, audio codec, dimensions, frame rate, and faststart metadata explicitly.
5. For destructive edits, write to a new output file and never overwrite the source unless the user explicitly asks.
## Common Commands
- Analyze: `ffprobe -hide_banner -show_streams -show_format input.mp4`
- Trim copy: `ffmpeg -ss 00:00:05 -to 00:00:20 -i input.mp4 -c copy output.mp4`
- H.264 web MP4: `ffmpeg -i input.mov -c:v libx264 -pix_fmt yuv420p -crf 20 -preset medium -c:a aac -b:a 160k -movflags +faststart output.mp4`
- Extract audio: `ffmpeg -i input.mp4 -vn -c:a pcm_s16le output.wav`
On macOS install system FFmpeg with `brew install ffmpeg`; on Windows use `winget install Gyan.FFmpeg`.
@@ -0,0 +1,29 @@
---
name: hyperframes
description: HyperFrames AI video generation workflow for turning a prompt, launch story, explainer script, product update, or motion graphics brief into deterministic HTML/CSS/JS animation rendered as MP4. Use when the user asks for text-to-video, product release videos, explainers, UI motion graphics, or agent-authored animated videos with HeyGen HyperFrames.
---
# HyperFrames Skill
Use this skill when the user wants an AI-generated short video where an agent writes HTML, CSS, and JavaScript animation and renders it deterministically to MP4.
Upstream full skill: https://github.com/heygen-com/hyperframes
Docs: https://hyperframes.heygen.com
Install full bundle: `npx skills add heygen-com/hyperframes` or `npx skills add heygen-com/hyperframes --all`
## Workflow
1. Clarify target duration, audience, aspect ratio, tone, brand constraints, and required call to action.
2. Convert the request into a compact video plan: scenes, timing, on-screen text, motion direction, and visual system.
3. Prefer deterministic HTML/CSS/JS animation over raster-only generation when text clarity, brand control, or repeatable rendering matters.
4. Keep each scene simple enough to render cleanly: one core idea, one motion focus, and short text.
5. Export or hand off a renderable project and call out where HyperFrames should render the final MP4.
## Good Fits
- Product launch or release-note videos.
- Explainer videos with diagrams, UI callouts, or animated typography.
- SaaS feature walkthroughs and marketing motion graphics.
- Social clips that need consistent brand text and layout.
Install the upstream HyperFrames skill for renderer commands, templates, and the complete 19-skill pack.
@@ -0,0 +1,27 @@
---
name: knowledge-cat-ppt
description: Knowledge Cat PPT workflow for knowledge-base-driven Chinese presentation creation, structured slide generation, teaching decks, reports, summaries, and PPTX refinement. Use when the user asks to turn documents, notes, knowledge cards, or research into a polished PowerPoint deck.
---
# Knowledge Cat PPT Skill
Use this skill when the user wants to turn knowledge, notes, or documents into a structured deck.
Upstream full skill: https://github.com/gnipbao/knowledge-cat-ppt-skill
## Workflow
1. Identify the deck goal, audience, language, duration, page count, and source materials.
2. Extract the source into a hierarchy: topic, subtopics, arguments, examples, data, and conclusion.
3. Build a slide outline before drafting slide content.
4. Keep each slide to one point, one visual idea, and presentation-ready wording.
5. Check consistency, text overflow, title hierarchy, and logical flow before final export.
## Good Fits
- Knowledge-base summaries into PPT.
- Training and teaching decks.
- Research notes into executive presentations.
- Chinese reports, courseware, and structured explanation decks.
Install `gnipbao/knowledge-cat-ppt-skill` from the Skills marketplace for the full implementation.
@@ -0,0 +1,28 @@
---
name: manim-animation
description: Manim animation workflow for math, science, education, explainer, and 3Blue1Brown-style knowledge videos. Use when the user asks for mathematical animations, diagrams, step-by-step visual proofs, concept explanations, or high-quality programmatic educational videos.
---
# Manim Animation Skill
Use this skill when the video should explain a concept through programmatic animation.
Manim Community: https://www.manim.community
Recommended skill repo: https://github.com/adithya-s-k/manim_skill
## Workflow
1. Convert the concept into a short storyboard: definitions, visual objects, transformations, and narration beats.
2. Keep each scene focused on one mathematical or conceptual transition.
3. Use Manim primitives for equations, axes, graphs, geometric objects, arrows, highlights, and camera motion.
4. Prefer clean timing, readable labels, and stepwise reveals over excessive effects.
5. Render short previews before final render when changing layout, camera, or equation density.
## Good Fits
- Math and science explainers.
- 3Blue1Brown-style animations.
- Diagrammatic product or architecture explanations.
- Teaching clips where visual transformation is the core value.
Install `adithya-s-k/manim_skill/skills/manimce-best-practices` for full ManimCE conventions.
@@ -0,0 +1,27 @@
---
name: openmontage
description: OpenMontage agentic video production workflow for intelligent editing, automatic montage assembly, vlog creation, research-to-script-to-assets-to-edit pipelines, and multi-agent video production. Use when the user asks to turn raw footage, references, or a topic into an edited video using OpenMontage-style pipelines.
---
# OpenMontage Skill
Use this skill when the user wants an agentic editing pipeline rather than a single render command.
Upstream project: https://github.com/calesthio/OpenMontage
## Workflow
1. Inventory available materials: raw footage, images, scripts, voiceover, music, brand assets, and target platform.
2. Build a pipeline plan: research, script, asset selection/generation, rough cut, captions, music, color, final render.
3. Split long videos into scenes or beats with clear inputs and outputs for each step.
4. Use FFmpeg or the project’s tools for deterministic trimming, concat, audio mixing, subtitle burn-in, and export.
5. Keep an edit decision list with source file, in/out time, reason, caption text, and transition notes.
## Good Fits
- Vlogs and recap videos from many clips.
- Topic-to-video production where research and script are part of the job.
- Short-form social edits needing subtitles, pacing, and overlays.
- Multi-asset projects where the user wants an AI editing assistant.
Install the upstream OpenMontage repository for its full 12-pipeline / tool system.
@@ -0,0 +1,29 @@
---
name: remotion-video
description: Remotion code-driven video workflow for React-based videos, reusable templates, dynamic props, batch rendering, data-driven motion graphics, product explainers, and programmatic video generation. Use when the user asks to create or render videos with React, Remotion, templates, captions, charts, or repeatable variants.
---
# Remotion Video Skill
Use this skill when the output should be a maintainable React video project or a batch-renderable template.
Remotion: https://www.remotion.dev
Repo: https://github.com/remotion-dev/remotion
Skills: `npx skills add remotion-dev/skills`
## Workflow
1. Define composition size, FPS, duration, and props schema before coding.
2. Create reusable scene components with predictable timing and layout.
3. Put user-editable text, colors, media paths, and data in props rather than hardcoding.
4. Use Remotion for deterministic charts, captions, UI walkthroughs, and templated social variants.
5. Validate render constraints: text overflow, safe areas, asset paths, audio duration, and final codec.
## Good Fits
- Product demos and UI motion.
- Data-driven video variants.
- Branded social clips with repeatable templates.
- Developer-controlled motion graphics.
Install the upstream Remotion skills for full project scaffolding and render commands.
@@ -0,0 +1,27 @@
---
name: video-use
description: Video-Use workflow for automatic raw footage editing, multi-source video assembly, image/text-to-video composition, filler-word removal, subtitles, color grading, and final.mp4 generation for coding agents. Use when the user gives footage or assets and asks for an edited video, montage, social clip, or creator-ready final.
---
# Video-Use Skill
Use this skill when the user wants to drop raw media into a folder and have an agent produce a polished final video.
Upstream project: https://github.com/browser-use/video-use
## Workflow
1. Ask for or locate the project media folder and target output format.
2. Inventory clips, audio, images, transcripts, and any reference examples.
3. Produce an edit plan: selects, cuts, captions, overlays, color, music, and final duration.
4. Use deterministic media tools such as FFmpeg for trims, concat, subtitles, audio, and export.
5. When animation overlays are needed, combine with HyperFrames, Remotion, or Manim depending on the visual style.
## Good Fits
- Creator clips from raw footage.
- Product demos assembled from screen recordings and overlays.
- Multi-material social videos with captions and music.
- “Here is a folder, make final.mp4” workflows.
Install the upstream Video-Use project for its full coding-agent media pipeline.
+98
View File
@@ -69,6 +69,69 @@ const BUILTIN_MARKETPLACE_SKILLS: &[BuiltinMarketplaceSkill] = &[
keywords: &["ppt", "powerpoint", "slides", "演示", "幻灯片", "汇报", "boosapi"],
fallback_skill_name: None,
},
BuiltinMarketplaceSkill {
name: "hyperframes",
description: "HyperFrames 文本进成片出:Agent 编写 HTML/CSS/JS 动画并确定性渲染 MP4,适合发布视频、explainers 和 motion graphics。",
repo: "heygen-com/hyperframes",
stars: 0,
installs: 0,
keywords: &["hyperframes", "heygen", "video", "mp4", "motion", "explainer", "html", "animation", "视频", "成片"],
fallback_skill_name: None,
},
BuiltinMarketplaceSkill {
name: "ffmpeg-media",
description: "FFmpeg 专业视频/音频处理 Skill:分析、压缩、转码、剪辑、拼接、字幕、音频提取和批处理命令生成。",
repo: "fabriqaai/ffmpeg-analyse-video-skill",
stars: 0,
installs: 0,
keywords: &["ffmpeg", "ffprobe", "video", "audio", "compress", "transcode", "subtitle", "剪辑", "压缩"],
fallback_skill_name: Some("ffmpeg-media"),
},
BuiltinMarketplaceSkill {
name: "openmontage",
description: "OpenMontage agentic 视频生产系统包装技能,用于智能剪辑、自动拼接、Vlog、研究到脚本到资产到剪辑的流水线。",
repo: "calesthio/OpenMontage",
stars: 0,
installs: 0,
keywords: &["openmontage", "video", "agentic", "vlog", "editing", "montage", "pipeline", "智能剪辑"],
fallback_skill_name: None,
},
BuiltinMarketplaceSkill {
name: "remotion-video",
description: "Remotion React 代码驱动视频 Skill:模板化批量出片、动态 props、数据视频、产品演示和 motion graphics。",
repo: "remotion-dev/skills",
stars: 0,
installs: 0,
keywords: &["remotion", "react", "video", "template", "motion", "render", "batch", "代码视频"],
fallback_skill_name: Some("remotion-video"),
},
BuiltinMarketplaceSkill {
name: "video-use",
description: "Video-Use 素材混剪 Skill:raw footage 到 final.mp4,支持素材组合、字幕、去口癖、color grading 和动画叠加规划。",
repo: "browser-use/video-use",
stars: 0,
installs: 0,
keywords: &["video-use", "browser-use", "footage", "final.mp4", "subtitle", "montage", "素材", "混剪"],
fallback_skill_name: None,
},
BuiltinMarketplaceSkill {
name: "manim-animation",
description: "Manim 数学/知识动画 Skill:3Blue1Brown 风格、公式、图形、概念解释、教学视频和程序化动画。",
repo: "adithya-s-k/manim_skill",
stars: 0,
installs: 0,
keywords: &["manim", "math", "animation", "3blue1brown", "education", "数学", "知识动画", "教学"],
fallback_skill_name: Some("manim-animation"),
},
BuiltinMarketplaceSkill {
name: "knowledge-cat-ppt",
description: "Knowledge Cat PPT Skill:把知识库、文档、笔记和研究材料整理成结构化中文 PPT / PowerPoint。",
repo: "gnipbao/knowledge-cat-ppt-skill",
stars: 0,
installs: 0,
keywords: &["knowledge-cat", "ppt", "powerpoint", "slides", "knowledge", "文档", "知识库", "课件"],
fallback_skill_name: None,
},
];
fn home_dir() -> PathBuf {
@@ -116,6 +179,41 @@ const BUNDLED_SKILLS: &[BundledSkill] = &[
source_ref: "GordenSun/GordenPPTSkill",
body: include_str!("../../builtin-skills/gorden-ppt-skill/SKILL.md"),
},
BundledSkill {
name: "hyperframes",
source_ref: "heygen-com/hyperframes",
body: include_str!("../../builtin-skills/hyperframes/SKILL.md"),
},
BundledSkill {
name: "ffmpeg-media",
source_ref: "fabriqaai/ffmpeg-analyse-video-skill",
body: include_str!("../../builtin-skills/ffmpeg-media/SKILL.md"),
},
BundledSkill {
name: "openmontage",
source_ref: "calesthio/OpenMontage",
body: include_str!("../../builtin-skills/openmontage/SKILL.md"),
},
BundledSkill {
name: "remotion-video",
source_ref: "remotion-dev/skills",
body: include_str!("../../builtin-skills/remotion-video/SKILL.md"),
},
BundledSkill {
name: "video-use",
source_ref: "browser-use/video-use",
body: include_str!("../../builtin-skills/video-use/SKILL.md"),
},
BundledSkill {
name: "manim-animation",
source_ref: "adithya-s-k/manim_skill",
body: include_str!("../../builtin-skills/manim-animation/SKILL.md"),
},
BundledSkill {
name: "knowledge-cat-ppt",
source_ref: "gnipbao/knowledge-cat-ppt-skill",
body: include_str!("../../builtin-skills/knowledge-cat-ppt/SKILL.md"),
},
];
pub fn seed_builtin_skills() -> Result<(), String> {