Skip to content

Generative Media Navigator Agent

The Generative Media Navigator is a Claude Code agent that automatically detects when you want to create videos or images, then invokes the /generative skill for you.

Skill + Agent Architecture

Component Purpose Type Install
/generative Skill Execute generation commands Skill ✅ Required
Generative Media Navigator Auto-detect intent, invoke skill Agent ✅ Recommended

Recommended Setup

Install both the Skill and Agent for the complete experience. The agent provides natural language understanding on top of the skill's execution capabilities.

Dependency

The agent requires the /generative skill to be installed first. See Skill Installation.

How It Works

User: "Create a video of a cat walking on the moon"
  ↓
Agent detects: video intent, prompt = "cat walking on the moon"
  ↓
Agent invokes: /generative video "cat walking on the moon"
  ↓
Result returned to user

Intent Detection

The agent recognizes these patterns:

Video Keywords:

Trigger Phrase Intent
create video, make video Video generation
I want a video, generate video Video generation
make a clip, create animation Video generation

Image Keywords:

Trigger Phrase Intent
create image, make image Image generation
I want an image, generate image Image generation
create a picture, make a photo Image generation

Installation

Prerequisites

  • ✅ Claude Code installed
  • ✅ /generative skill installed (see Installation)

Linux / macOS

# Create agents directory
mkdir -p ~/.claude/agents

# Copy agent file from repository
cp .claude/agents/generative-media-navigator.md ~/.claude/agents/

# Restart Claude Code

Windows (PowerShell)

# Create agents directory
New-Item -ItemType Directory -Force -Path "$env:USERPROFILE\.claude\agents"

# Copy agent file
Copy-Item ".\.claude\agents\generative-media-navigator.md" `
   -Destination "$env:USERPROFILE\.claude\agents\"

# Restart Claude Code

Verification

After installation, test with natural language:

You: "Create a video of a cat walking on the moon"
Agent: [Detects video intent → invokes /generative video "cat walking on the moon"]

You: "Create an image of a beautiful sunset"
Agent: [Detects image intent → invokes /generative image "beautiful sunset"]

You: "Make a video showing ocean waves"
Agent: [Detects video intent → invokes /generative video "ocean waves"]

When Agent Does NOT Act

The agent will not auto-invoke when:

  • Asking about previous generated content ("How did the video I made earlier turn out?")
  • Troubleshooting or checking status
  • Request is ambiguous (will ask for clarification)
  • User explicitly says they don't want to generate

Directory Structure

After full installation (Skill + Agent):

~/.claude/
├── agents/
│   └── generative-media-navigator.md   ← Agent (enhancement)
└── skills/
    └── generative/
        └── SKILL.md                      ← Skill (required)

Agent Configuration

The agent uses these settings:

Setting Value Description
name generative-media-navigator Unique identifier
tools Bash, Read, Glob, Grep, AskUserQuestion, Skill Available tools
model inherit Uses parent model
color purple UI indicator color

Updating

# Pull latest changes
cd claude-code-media-generator
git pull

# Update agent file
# Linux/macOS:
cp .claude/agents/generative-media-navigator.md ~/.claude/agents/

# Windows (PowerShell):
Copy-Item ".\.claude\agents\generative-media-navigator.md" `
   -Destination "$env:USERPROFILE\.claude\agents\" -Force

# Restart Claude Code