Token导航 LogoToken导航

Agent Skill · 研究检索

shellbot-creative-studio(贝壳机器人创意工作室)

shellbot-creative-studio 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

shellbot-creative-studio 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为接触敏感数据;许可证为 MIT-0,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 24,581。

总安装 3,975周安装 169GitHub Stars 公开资料未说明下载 1,393核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
1.0.3

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:shellbot-creative-studio(贝壳机器人创意工作室)
来源仓库:https://github.com/cohnen/shellbot-creative-studio
安装命令:
openclaw skills install shellbot-creative-studio
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install shellbot-creative-studio

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

Shellbot-creative-studio 是面向 AI Agent 的无头创意制作工作室。支持图像生成、照片编辑和视频合成等多媒体处理能力。通过 clawhub 安装,安装命令为 openclaw skills install shellbot-creative-studio。可组合多种输出模块形成完整作品。

需配置渲染参数如分辨率与时长以适应不同用途。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install shellbot-creative-studio;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

shellbot-creative-studio 怎么安装?

在终端执行 openclaw skills install shellbot-creative-studio。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

shellbot-creative-studio 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

shellbot-creative-studio 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

shellbot-creative-studio 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

shellbot-creative-studio 用的人多吗?

公开数据是累计安装 3,975 次,近一周安装 169 次,下载 1,393 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

shellbot-creative-studio 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
shellbot-creative-studio

贝壳机器人创意工作室 · 本页

3,975—接触敏感数据MIT-0通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
shellbot-creative-studio
version
1.0.2
description
Headless creative production studio for AI agents. Generate images,
allowed-tools
Bash(curl *api.freepik.com*), Bash(curl **.freepik.com*), Bash(curl *queue.fal.run*),
argument-hint
<command> [--param value]
metadata
openclaw
emoji
🎬
primaryEnv
FREEPIK_API_KEY
providerEnv
requires
env
anyBins
homepage
https://getshell.ai
tags

shellbot-creative-studio

Headless creative production studio for AI agents. One skill to generate images, edit photos, create videos, produce audio, and assemble everything into polished video output via Remotion.

Philosophy: Opinionated defaults, narrow overrides. Just works with whatever API keys you have.

Commands

CommandDescriptionExample
imageGenerate images (T2I, reference-guided)image --prompt "product shot" --style photo
editEdit images (upscale, remove-bg, inpaint, outpaint, style-transfer, relight)edit --input img.png --action upscale
videoGenerate video clips (T2V, I2V)video --prompt "cinematic intro" --duration 5
voiceText-to-speech narrationvoice --text "Welcome to..." --output vo.mp3
musicBackground music generationmusic --prompt "upbeat cinematic" --duration 30
sfxSound effects generationsfx --prompt "whoosh transition" --duration 2
remotion initBootstrap Remotion project from templateremotion init --template cinematic-product-16x9
remotion renderRender Remotion project to MP4remotion render --project ./my-video
planCreative brief to storyboard JSONplan --brief "Launch video for..." --framework aida
pipelineEnd-to-end production (plan to render)pipeline --brief "..." --template aida-classic-16x9
statusCheck async task statusstatus --task-id abc123 --provider freepik
providersShow available providersproviders

Read references/command-reference.md for full parameter docs.

Arguments

  • Command: $0 — one of the commands above
  • All args: $ARGUMENTS — forwarded to the command script

Authentication

Set one or more of these environment variables:

VariableProviderCapabilitiesSetup
FREEPIK_API_KEYFreepikimage, edit, video, voice, music, sfxhttps://www.freepik.com/api/keys
FAL_API_KEYfal.ai (+ Nano Banana 2 via fal)image, edit, videohttps://fal.ai/dashboard/keys
GOOGLE_API_KEYNano Banana 2 (via Google)image, edithttps://aistudio.google.com/apikey
OPENROUTER_API_KEYOpenRouter (Nano Banana 2 + many image models)imagehttps://openrouter.ai/keys
ELEVENLABS_API_KEYElevenLabsvoice, music, sfxhttps://elevenlabs.io/app/settings/api-keys

Nano Banana 2 auto-detects the best available backend: GOOGLE_API_KEY (preferred) > FAL_API_KEY (fal-ai/nano-banana-2) > OPENROUTER_API_KEY > infsh CLI. OpenRouter also works as a standalone image provider with access to many models — see https://openrouter.ai/collections/image-models

At minimum, set FREEPIK_API_KEY (covers all capabilities). Run providers to check what's available.

Provider auto-routing

Each command auto-selects the best provider based on available API keys. Override with --provider.

TaskPriority 1Priority 2Priority 3
imageNano Banana 2 (GOOGLE_API_KEY / FAL_API_KEY / OPENROUTER_API_KEY)Freepik Seedream v5 litefal Flux-2 / OpenRouter
editFreepik (precision upscale)fal (creative upscale)Nano Banana 2 (iterative)
videoFreepik Kling v3 (quality)fal Kling v2 (fallback)—
voiceFreepik ElevenLabsElevenLabs direct—
musicFreepik music genElevenLabs—
sfxFreepik SFXElevenLabs—

Read references/provider-matrix.md for the full model selection decision tree.


Opinionated creative workflow

This is how to think about when to use each tool. Follow this decision framework.

When to use Nano Banana 2 (Gemini) for images

Nano Banana 2 is the default for image generation because:

  • Best multi-image consistency (same character, same product across scenes)
  • Instruction-based editing with up to 14 reference images
  • Native Google Search grounding for real-world accuracy
  • Fast and cheap (~$0.067/1K image on Flash, ~$0.134 on Pro)

Use Nano Banana 2 (gemini-3.1-flash-image-preview) when:

  • Ideating scene visuals (fast iteration, batch 4 variants)
  • Building a visual kit (hero, alt angle, detail, background — same style)
  • Editing existing images with text instructions
  • Combining multiple reference images into a new composition

Use Nano Banana Pro (gemini-3-pro-image-preview) when:

  • Maximum quality needed for hero shots
  • Complex compositions with precise text rendering
  • Final delivery assets (not just concepting)

Switch to Freepik Seedream v5 lite when:

  • Fast, high-quality generation with Freepik (default Freepik model)
  • You want Freepik's rendering quality without specifying a model
  • Commercial-ready output

Other Freepik models (use --model):

  • seedream-v4-5 — typography-heavy posters and ad creatives
  • mystic — ultra-photoreal product photography (slower, higher cost)
  • flux-2-klein — sub-second generation for rapid concepting

When to use Kling (video) vs Remotion

This is the most important decision. Never use AI video for what Remotion does better.

Use Kling (video command) for:

  • Hero moments: product reveals, cinematic openings, emotional beats (3-8s clips)
  • Organic motion that's hard to code: water, fire, fabric, hair, camera orbits
  • Image-to-video: animate a still into a living scene
  • Transition inserts: short clips between Remotion scenes

Use Remotion for:

  • Text animations, captions, kinetic typography
  • Precise timing synced to voiceover
  • Brand overlays, logos, consistent color grading
  • Data visualizations, metric counters, charts
  • Scene transitions (cuts, wipes, dissolves — deterministic)
  • Final assembly: compositing Kling clips + stills + audio + text

The ideal workflow:

  1. Generate scene stills with Nano Banana 2 (consistency across scenes)
  2. Animate 2-3 hero stills with Kling (cinematic motion)
  3. Generate voiceover + music
  4. Assemble everything in Remotion (timing, text, transitions, audio mix)

When to use the pipeline command vs manual steps

Use pipeline for standard product marketing or explainer videos where you trust the defaults. It runs the full plan-to-render sequence automatically.

Use manual steps when:

  • You need to review/cherry-pick assets between steps
  • The creative direction requires unusual choices
  • You want to iterate on individual scenes
  • You're composing a complex multi-format campaign (16:9 + 9:16 + stills)

Read references/prompt-cookbook.md for proven prompt patterns for every command — product photography, SaaS UI, transparent assets, video reveals, audio recipes, and full end-to-end examples.


Command behavior

$0 = image

Generate images with smart provider selection.

bash scripts/image.sh --prompt "Studio product shot of wireless earbuds, soft lighting" \
  --style photo --size 16:9 --resolution 2k --output hero.png

Force a specific provider or model:

bash scripts/image.sh --prompt "..." --provider freepik --model seedream-v4-5
bash scripts/image.sh --prompt "..." --provider nano-banana-2 --count 4
bash scripts/image.sh --prompt "..." --provider nano-banana-2 --model pro  # Nano Banana Pro

Nano Banana 2 models:

  • gemini-3.1-flash-image-preview (default) — fast, cheap, great consistency
  • pro or gemini-3-pro-image-preview — highest quality, 2x cost

$0 = edit

Edit images with 6 actions: upscale, remove-bg, inpaint, outpaint, style-transfer, relight.

bash scripts/edit.sh --input hero.png --action upscale --scale 4 --output hero-4k.png
bash scripts/edit.sh --input photo.jpg --action remove-bg --output photo-nobg.png
bash scripts/edit.sh --input scene.png --action outpaint --left 512 --right 512 --prompt "extend the landscape"

$0 = video

Generate video clips. Supports text-to-video and image-to-video.

Opinionated approach: Always generate a still first with image, review it, then animate with video --image. This gives you control over the starting frame.

bash scripts/video.sh --prompt "Product floating in space, rotating slowly" \
  --duration 5 --aspect 16:9 --output hero-clip.mp4

bash scripts/video.sh --prompt "Camera orbits around the product" \
  --image hero.png --duration 5 --output orbit-clip.mp4

$0 = voice

Generate voiceover narration.

bash scripts/voice.sh --text "Welcome to the future of productivity." \
  --voice 21m00Tcm4TlvDq8ikWAM --output vo-intro.mp3

$0 = music

Generate background music.

bash scripts/music.sh --prompt "cinematic ambient, subtle and professional" \
  --duration 45 --output music.mp3

$0 = sfx

Generate sound effects.

bash scripts/sfx.sh --prompt "soft whoosh transition" --duration 1.5 --output whoosh.mp3

$0 = remotion init

Bootstrap a Remotion project from bundled templates.

bash scripts/remotion_init.sh --template cinematic-product-16x9 --name MyProduct --output ./my-video
bash scripts/remotion_init.sh --list  # Show available templates

Templates: aida-classic-16x9, cinematic-product-16x9, saas-metrics-16x9, mobile-ugc-9x16, blank-16x9, explainer-16x9

All templates use Remotion 4.0.419 (pinned), React 18, TypeScript. Each includes npm run start (studio), npm run render, and npm run verify.

$0 = remotion render

Render a Remotion project to MP4.

bash scripts/remotion_render.sh --project ./my-video --output final.mp4 --codec h264 --crf 18

$0 = plan

Convert a creative brief into a structured storyboard.

bash scripts/plan.sh --brief "Launch video for FitPulse, a fitness app for busy professionals" \
  --framework aida --duration 45 --output storyboard.json

For AIDA framework, provide structured brief JSON:

bash scripts/plan.sh --brief-file brief.json --framework aida --output plan.json

Required AIDA fields: product_name, audience, problem, solution, use_cases (array), cta, incentive

$0 = pipeline

Run the full production pipeline: brief to rendered video.

bash scripts/pipeline.sh \
  --brief "Launch video for a new AI writing assistant targeting content creators" \
  --template cinematic-product-16x9 \
  --output-dir ./creative-output \
  --duration 45

# Dry run to preview the plan:
bash scripts/pipeline.sh --brief "..." --dry-run

# Stop before render for review:
bash scripts/pipeline.sh --brief "..." --skip-render

Pipeline steps: plan > generate images > upscale heroes > generate video clips > voiceover > music > Remotion assembly > render

$0 = providers

Check which providers are available.

bash scripts/providers.sh

$0 = status

Check async task status.

bash scripts/status.sh --task-id abc123 --provider freepik
bash scripts/status.sh --all  # Show all tracked tasks

Output convention

  • JSON to stdout — machine-readable, pipe-friendly
  • Colored logs to stderr — human-readable status updates
  • All commands support --dry-run to preview without API calls
  • Async tasks are tracked in .creative-tasks.jsonl per output directory

Remotion principles

  • Remotion is the source of truth for timing, layout, animation, and render
  • External tools (Freepik, ElevenLabs, Kling) generate ancillary assets only
  • Transitions: 8-18 frames, purposeful (not decorative)
  • Load fonts explicitly with @remotion/google-fonts
  • Always run npm run verify before npm run render
  • Load reference rules from references/remotion-rules/ as needed

Read references/remotion-playbook.md for detailed Remotion implementation guidance.

Local resources

Scripts (CLI commands)

  • scripts/creative.sh — Main router (dispatches to sub-commands)
  • scripts/image.sh — Image generation
  • scripts/edit.sh — Image editing (6 actions)
  • scripts/video.sh — Video generation
  • scripts/voice.sh — Text-to-speech
  • scripts/music.sh — Music generation
  • scripts/sfx.sh — Sound effects
  • scripts/remotion_init.sh — Remotion project bootstrap
  • scripts/remotion_render.sh — Remotion render
  • scripts/plan.sh — Brief to storyboard
  • scripts/pipeline.sh — End-to-end pipeline
  • scripts/status.sh — Task status checker
  • scripts/providers.sh — Provider availability

References

  • references/command-reference.md — Full parameter docs for every command
  • references/provider-matrix.md — Provider/model decision tree
  • references/creative-guidelines.md — Quality standards
  • references/workflow-recipes.md — End-to-end recipe examples
  • references/remotion-playbook.md — Remotion implementation guide
  • references/freepik-ancillary-assets.md — Asset generation recipes
  • references/template-showcase.md — Template selector guide
  • references/remotion-rules-index.md — Index of 29+ Remotion rule files
  • references/remotion-rules/ — Detailed Remotion rules (animations, audio, text, transitions, etc.)

Templates (Remotion projects)

  • assets/templates/aida-classic-16x9/ — AIDA product marketing
  • assets/templates/cinematic-product-16x9/ — Dramatic product launch
  • assets/templates/saas-metrics-16x9/ — B2B dashboard metrics
  • assets/templates/mobile-ugc-9x16/ — Vertical social (Reels/TikTok)
  • assets/templates/blank-16x9/ — Minimal starter
  • assets/templates/explainer-16x9/ — 5-scene how-it-works

适合场景

01

调用多模型

02

代码和文本生成

03

Agent 推理流程

04

OpenRouter 模型接入

能力概览

能力 1

统一调用多种 LLM

能力 2

支持 Claude、Gemini、Kimi 等模型

能力 3

适合聊天、代码和推理任务

能力 4

可作为 Agent 模型调用入口

平台分布

OpenClaw

81.17%
按下载量换算1,131

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills