Skip to content

gitroomhq/agent-media

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

22 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

agent-media skill

A Claude Code skill that drives the agent-media CLI to generate AI UGC videos — Selfie clips, reusable characters, burned-in subtitles — from natural-language prompts inside Claude Code, Cursor, Continue, or any MCP-aware AI client.

Install (Claude Code)

# 1. Install the CLI + log in
npm install -g agent-media-cli@latest
agent-media login

# 2. Add the skill
claude skills add github:gitroomhq/agent-media

Restart Claude Code, then describe what you want — the skill walks you through 4 gates (script, character, director's brief, duration), runs the CLI, and returns the mp4 URL.

Install (Claude.ai / Cowork / Claude Desktop)

These use a hosted HTTP MCP endpoint:

  1. Settings → Integrations → Add custom MCP server
  2. URL: https://api.agent-media.ai/mcp
  3. Auth header: Authorization: Bearer <your-ma_xxx-key>

Tools create_selfie, create_character, create_subtitle show up in the picker.

Install (Cursor / Continue / Windsurf)

Add to ~/.cursor/mcp.json (or your client's equivalent):

{
  "mcpServers": {
    "agent-media": {
      "command": "npx",
      "args": ["-y", "@agentmedia/mcp-server"],
      "env": { "AGENT_MEDIA_API_KEY": "<your-ma_xxx-key>" }
    }
  }
}

Keeping the skill up-to-date

agent-media skill update       # pulls the latest SKILL.md + reference/*
agent-media skill status       # local vs remote version

Every CLI invocation runs a once-per-day background check and prints a one-line nudge when a newer skill version is available. Opt-in silent auto-update with AGENT_MEDIA_AUTO_UPDATE_SKILL=1; opt-out with AGENT_MEDIA_SKIP_SKILL_CHECK=1.

What this repo ships

SKILL.md                         # ~4 KB index + hard-gate directive
reference/
  conversation-flow.md           # MUST READ — the 4 gates protocol
  presets.md                     # 20 shot presets + "best for" hints
  vibes.md                       # 5 vibes (excited, calm, sassy, …)
  subtitle-styles.md             # 17 subtitle styles
  realism-rubric.md              # 9-point visual-quality guard
  errors.md                      # common error codes + fixes
  generators/
    selfie.md                    # talking-head pipeline deep dive
    character_create.md          # persisted reusable character
    subtitle.md                  # burn subs onto any video
docs/
  api-reference.md               # auto-generated REST reference

What the skill does

Generator What it makes
Selfie 9:16 TikTok-style video. Talking-head AI person, 5/10/15 s, 20 shot presets, 5 vibes, 17 subtitle styles. No photo required — generates the character from a text description.
Character Persists a reusable character so subsequent Selfies stay on-model.
Subtitle Burns styled subs onto any existing video. 17 styles, optional transcript override.

Pricing

Pricing lives at https://agent-media.ai/pricing. The API debits internally — agents and clients should not surface credit numbers or USD to end users.

Companion packages

Links

Apache-2.0 · 2026 agent-media contributors

About

Generate AI-powered videos and images from the terminal using the `agent-media` CLI.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors