video-cog

Long-form AI video production: the frontier of multi-agent coordination. CellCog orchestrates 6-7 foundation models to produce up to 4-minute videos from a single prompt — scripted, filmed, voiced, lipsync'd, scored, and edited automatically. Create marketing videos, product demos, explainer videos, educational content, spokesperson videos, training materials, UGC content, news reports.

Security Scan

The skill is an instruction-only wrapper that delegates heavy work to an external 'cellcog' skill — the instructions are mostly coherent with the description, but important pieces (the dependency's source, credentials, and runtime behavior) are not visible here and could introduce unexpected access or network activity.

Purpose & Capability

The name/description (multi-model video production) match the runtime instructions: the SKILL.md explicitly delegates all heavy lifting to the 'cellcog' skill and shows example SDK usage. That delegation is coherent, but the skill itself contains no implementation — it only instructs the agent to use another skill. Minor inconsistency: the doc mentions both 'up to 4-minute videos' and elsewhere '15 seconds to 5+ minutes'.

Instruction Scope

SKILL.md stays on-topic: it instructs installing and using the 'cellcog' skill and how to call client.create_chat with agent-team mode and notify_session_key. It does not instruct the agent to read arbitrary files or environment variables. However, it explicitly requires using an 'agent team' multi-agent mode (which can trigger many downstream actions) and instructs not to poll but to rely on a daemon notification — this hands considerable runtime discretion to the dependency.

Install Mechanism

There is no install spec or code in this skill itself (lowest direct install risk). The SKILL.md tells users to 'clawhub install cellcog' — installing that dependency is required and transfers risk to whatever that package contains. Because the dependency's source is not visible here, you should review the cellcog skill before installation.

Credentials

This skill declares no required env vars or credentials itself, but it depends entirely on 'cellcog' for SDK setup and API calls — that dependency will likely require API keys/credentials (model provider keys, storage, etc.). The SKILL.md references a notify_session_key value format; although the skill doesn't request secrets directly, installing/using the dependency will probably require sensitive credentials. The lack of explicit credential declaration here means credential requests will come from the dependency and are not visible in this manifest.

Persistence & Privilege

The skill does not set always:true and does not disable model invocation, so the agent can invoke it autonomously when eligible. Because it delegates to an external multi-agent workflow, enabling autonomous invocation could cause background network activity, model calls, or use of credentials supplied to the dependency. If you want to limit autonomous runs, consider disabling model invocation or requiring explicit user confirmation before invoking the dependency.

What to consider before installing

This skill is effectively a set of instructions that tell your agent to call another skill ('cellcog') to do multi-model video production. That makes it low-risk by itself (no code, no env vars), but the real risk is the dependency you will install. Before installing or using this skill: 1) Inspect the 'cellcog' skill source and homepage — confirm its publisher, code, and install steps; 2) Check what credentials (API keys, storage, model provider tokens) 'cellcog' requires and whether those are appropriate to provide; 3) Prefer installing from a known, auditable source (GitHub release or official registry) and review any extracted/installed binaries; 4) If you want to limit surprise activity, set disableModelInvocation:true on the skill or require explicit user confirmation before the agent runs long-running multi-agent jobs; 5) Be aware of the minor doc inconsistency about maximum duration (4 min vs 5+ min). If you cannot review the 'cellcog' dependency, treat this skill as higher risk and avoid installing it.

Video Cog - The Frontier of Multi-Agent Video Production

Long-form AI video production is the hardest challenge in multi-agent coordination. CellCog may be the only platform that pulls it off.

6-7 foundation models orchestrated to produce up to 4-minute videos from a single prompt: script writing, scene generation, voice synthesis, lipsync, music scoring, and editing — all automatic. Marketing videos, product demos, explainers, educational content, AI spokesperson videos, UGC, news reports, and more.

---

Prerequisites

This skill requires the cellcog skill for SDK setup and API calls.

clawhub install cellcog

Read the cellcog skill first for SDK setup. This skill shows you what's possible.

Quick pattern (v1.0+):

# Fire-and-forget - returns immediately
result = client.create_chat(
    prompt="[your video request]",
    notify_session_key="agent:main:main",
    task_label="video-task",
    chat_mode="agent team"
)
# Daemon notifies you when complete - do NOT poll

---

What Videos You Can Create

Marketing Videos

Promotional content for products and services:

• Product Demos: "Create a 30-second product demo video for our new fitness app showing key features"
• Brand Videos: "Generate a 60-second brand story video for an eco-friendly clothing company"
• Social Ads: "Create a 15-second Instagram ad for a coffee subscription service"
• Launch Videos: "Make a product launch announcement video for a new AI writing tool"

Explainer Videos

Educational content that breaks down complex topics:

• Product Explainers: "Create an explainer video showing how our SaaS platform works"
• Concept Explanations: "Make a video explaining how blockchain works for beginners"
• Process Walkthroughs: "Generate a video explaining the mortgage application process"
• Feature Tours: "Create a video tour of our app's new dashboard features"

Educational Videos

Learning content for courses and training:

• Tutorial Videos: "Create a tutorial video on Python list comprehensions"
• Course Content: "Generate a lesson video on the causes of World War I"
• Training Materials: "Make an employee onboarding video about our company values"
• How-To Guides: "Create a how-to video for setting up a home studio for podcasting"

Documentary Style

Informative, story-driven content:

• Mini Documentaries: "Create a 3-minute documentary-style video about the rise of electric vehicles"
• Company Stories: "Generate a documentary about our startup journey"
• Industry Deep Dives: "Make a documentary exploring the future of space tourism"
• Historical Content: "Create a documentary-style video about the history of Silicon Valley"

Cinematic / Creative

Artistic and visually striking content:

• Short Films: "Create a 2-minute cinematic short about a day in Tokyo"
• Mood Pieces: "Generate a cinematic video capturing the energy of a busy coffee shop"
• Music Video Style: "Create a visually dynamic video for an electronic music track"
• Artistic Showcases: "Make a cinematic portfolio video for a photographer"

UGC (User Generated Content) Style

Authentic, relatable content that feels personal:

• Testimonial Style: "Create a UGC-style testimonial video for a skincare product"
• Unboxing Style: "Generate an unboxing-style video for a new tech gadget"
• Day-in-the-Life: "Make a day-in-the-life style video featuring a remote worker using our app"
• Review Style: "Create a casual review-style video for a meal delivery service"

News / Reporting Style

Professional news-format content:

• News Reports: "Create a news-style report video about the latest AI developments"
• Market Updates: "Generate a financial news video about tech stock earnings"
• Industry News: "Make a news report about new regulations in the fintech space"
• Analysis Pieces: "Create a news analysis video about the state of remote work"

---

Lipsync & Spokesperson Videos

CellCog can generate videos with AI characters speaking your script:

• AI Spokesperson: "Create a video with a professional spokesperson explaining our product"
• Avatar Presentations: "Generate a video with an AI presenter delivering our quarterly update"
• Character Narration: "Make a video with a friendly character explaining our children's app"

For lipsync videos:

1. The starting frame should show only one human face prominently
2. Provide the script/dialogue
3. CellCog handles voice synthesis and lip synchronization

---

Video Specifications

Aspect	Options
Duration	15 seconds to 5+ minutes
Aspect Ratios	16:9 (landscape), 9:16 (portrait/mobile), 1:1 (square)
Styles	Photorealistic, animated, cinematic, documentary, casual
Audio	Background music, voiceover, sound effects, or silent

---

When to Use Agent Team Mode

For video generation, always use chat_mode="agent team" (the default).

Video creation involves:

• Script writing
• Scene planning
• Image generation for frames
• Audio generation
• Video synthesis
• Quality review

This multi-step process requires the full agent team for best results.

---

Example Video Prompts

Marketing video:

"Create a 30-second marketing video for 'FreshBrew' - a premium coffee subscription. Show beautiful coffee preparation scenes, happy customers, and end with our tagline 'Freshness Delivered Daily'. Upbeat background music, no voiceover. 16:9 for YouTube."

Explainer with voiceover:

"Create a 90-second explainer video for our project management tool. Walk through: 1) Creating a project, 2) Adding team members, 3) Tracking progress. Professional female voiceover, clean animated style, include captions. 16:9 format."

Educational content:

"Generate a 3-minute educational video explaining photosynthesis for middle school students. Use engaging animations, clear narration, and include a summary at the end. Friendly, approachable style."

Spokesperson video:

"Create a 60-second video with an AI spokesperson (professional male, 30s) announcing our Series B funding. Script: 'Today, we're thrilled to announce...' [provide full script]. Business casual setting, confident tone."

---

Tips for Better Videos

1. Specify duration: "30 seconds" or "2 minutes" helps scope the content appropriately.

2. Define aspect ratio: 16:9 for YouTube/web, 9:16 for TikTok/Reels/Shorts, 1:1 for Instagram feed.

3. Describe the style: "Cinematic", "casual UGC", "corporate professional", "playful animated".

4. Audio preferences: "Upbeat music", "calm narration", "no audio", "sound effects only".

5. Include key moments: Describe the scenes or beats you want to hit.

6. Provide scripts: For spokesperson/voiceover videos, write out exactly what should be said.