This repository offers an Agent Skill for Agnes AI's text, image, and video generation APIs, enabling AI clients supporting the Agent Skills framework to directly leverage Agnes's multimedia capabilities. It targets users of platforms like Codex and Claude Code, providing a standardized integration for advanced AI-driven content creation.

How It Works

The skill encapsulates Agnes AI's official APIs within a standard SKILL.md structure, facilitating integration with various AI clients. It intercepts agent requests, automatically translates non-English prompts for improved video generation stability, and manages asynchronous video task creation and polling. This approach offers seamless integration and robust handling of complex media generation workflows.

Quick Start & Requirements

Installation requires npx skills add Yacey/agnes-ai-generation-skill. A valid Agnes API Key from https://platform.agnes.ai/ is mandatory, configured via the AGNES_API_KEY environment variable. The skill relies on Node.js/npx for installation and Python for script execution.

Highlighted Details

• Supports text generation (streamed, OpenAI-compatible tool calls).
• Advanced image generation (text-to-image, image-to-image) with high-information density and cinematic realism.
• Comprehensive video generation (text-to-video, image-to-video, multi-image, keyframes) with asynchronous task management, polling, and cinematic output.
• Automatic translation of non-English prompts for enhanced video generation stability.
• Media results are returned as URLs by default, optimizing for agent workflows.

Maintenance & Community

The project is at version v0.1.0, indicating early development. The README provides no details on specific contributors, community channels, or a public roadmap.

Licensing & Compatibility

Released under the MIT License, permitting commercial use and integration into closed-source applications. Compatible with AI clients supporting the Agent Skills framework.

Limitations & Caveats

Agnes Responses API's multi-turn function calls are unsuitable for automated agent loops. Tool calls are for request structure compatibility, not stable multi-turn execution. While video generation is supported, specific multi-image/keyframe modes may require individual testing. A past server-side "division by zero" error during video task querying was noted, though subsequent video tasks have succeeded.