Noureddine RAMDI / heygen-com/skills: modular AI agent skills for avatar and video generation

Created Mon, 04 May 2026 10:23:02 +0000 Modified Sat, 23 May 2026 20:41:27 +0000
877 Words
github-stars ai agent-skills avatar video-generation shell cli heygen

heygen-com/skills

Generating digital avatars and videos from photos is a common yet complex task in AI-powered multimedia systems. heygen-com/skills tackles this by packaging two modular AI agent skills that create digital twin avatars and generate videos with those avatars, all orchestrated through a pipeline of human/machine-readable state files. What makes this repo stand out is its agent-driven install pattern, where a single prompt instructs the AI agent to clone the repo, request API keys, pick the best transport, verify the setup, and produce a working video — all automated through the repo’s own install spec.

Modular AI agent skills for avatar and video generation

At its core, heygen-com/skills is a collection of two AI agent “skills”: heygen-avatar and heygen-video. These skills enable coding agents such as Claude Code, Cursor, Codex, OpenClaw, and others to create a digital avatar from a photo and then generate videos featuring that avatar.

The repo is written in shell scripts and structured to be compatible across multiple AI coding agents’ skill-loading conventions. The code is surprisingly lightweight — the HeyGen CLI is a single static binary with no runtime dependencies, simplifying deployment and usage.

Communication between the skills and the agents happens via a set of AVATAR-*.md state files. These files are both human- and machine-readable, allowing the avatar creation and video generation steps to be chained together in a pipeline:

  • A photo is converted into an avatar digital twin.
  • The avatar state is saved and can be inspected or modified.
  • Videos are generated using the avatar state.

Under the hood, the skills interact with the HeyGen API, which requires authentication either via API key (CLI mode) or OAuth tokens (MCP mode) that consume existing HeyGen plan credits. This dual authentication mechanism allows flexibility depending on the environment and usage scenario.

Agent-driven install pattern and integration tradeoffs

What truly distinguishes this repo is the install approach. Instead of manual setup, the repo embeds its installation instructions in a single prompt stored in INSTALL_FOR_AGENTS.md. When you paste this prompt into your AI coding agent, it automatically:

  • Clones the repo to the correct path.
  • Requests your HeyGen API key interactively.
  • Selects the appropriate transport mechanism.
  • Runs a verification test.
  • Ends with a working generated video.

This pattern shifts the install logic from the user’s clipboard or local instructions into the repo itself. It ensures the install process is reproducible, versioned, and agent-driven — a neat example of coder-agent synergy.

The tradeoff here is that this puts a lot of trust in the AI agent’s ability to interpret and execute the install prompt correctly. Also, the skills depend on HeyGen’s cloud API, so offline or self-hosted usage is not possible. The binary CLI has no runtime dependencies, which is good for portability but may limit extensibility.

The code quality is pragmatic with clear separation between the two skills and well-documented state file conventions. The repo doesn’t try to reinvent avatar or video generation algorithms but rather wraps the HeyGen API as reusable agent skills, focusing on integration and DX.

Install and quickstart with AI coding agents

The README provides several explicit installation options, reflecting the diversity of AI agent runtimes:

  • gh skill install commands for GitHub CLI v2.90+, supporting many agents like Claude Code, Cursor, Codex, and more:
gh skill install heygen-com/skills heygen-avatar
gh skill install heygen-com/skills heygen-video

  • clawhub install heygen-skills for ClawHub users.

  • openclaw plugins install clawhub:@heygen/openclaw-plugin-heygen for OpenClaw plugin users.

  • Manual git clone for agents that don’t support the above installers, with specific install paths per agent.

After installation, the two skills are discovered at heygen-avatar/SKILL.md and heygen-video/SKILL.md.

For authentication, you generate an API key from the HeyGen web app and provide it when prompted during install or skill use.

The key point is the single prompt install approach:

Read https://raw.githubusercontent.com/heygen-com/skills/master/INSTALL_FOR_AGENTS.md and follow it.
Ask me for any API keys you need.

Pasting that into your agent triggers the whole install and first video generation.

Who should consider heygen-com/skills?

This repo is relevant for developers building AI coding agents or multi-agent systems who want to add multimedia capabilities — specifically avatar creation and video generation — without building these complex pipelines from scratch.

Its agent-driven install pattern is worth understanding for anyone extending AI agents with reusable skill modules. The human- and machine-readable state file communication is a clean pattern for chaining multi-step AI workflows.

The limitation is the dependency on HeyGen’s cloud API and the assumption your agent environment supports the install paths and CLI binaries. It’s not a drop-in local avatar/video generator but rather a bridge to a cloud service wrapped as modular skills.

If you’re working on AI agent extensibility or integrating multimedia generation into agent workflows, this repo is a practical reference and a ready-to-use toolkit.

→ GitHub Repo: heygen-com/skills ⭐ 208 · Shell