fal-ai-skills

fal.ai Skills

Skills for fal.ai — discover, compare, and run 1000+ AI models for image, video, audio, and 3D generation directly from your coding agent.

How It Works

When you ask your agent to generate an image, create a video, or run any fal.ai model, the skill activates automatically. It guides the agent through a workflow:

1. Discover — search fal.ai's model catalog for candidates
2. Present options — show you models with pricing, let you pick
3. Probe — run a cheap sample to validate the approach
4. Refine — adjust based on your feedback, re-probe as needed
5. Execute — run at full quality once you're happy

The agent never picks a model or runs anything without your approval.

Installation

Claude Code

# Add the marketplace
claude plugin marketplace add morozovdd/fal-ai-skill

# Install the plugin
claude plugin install fal-ai

Cursor

Search for "fal-ai" in the plugin marketplace, or install from .cursor-plugin/plugin.json.

Codex

Tell Codex:

Fetch and follow instructions from https://raw.githubusercontent.com/morozovdd/fal-ai-skill/main/.codex/INSTALL.md

GitHub Copilot CLI

copilot plugin add morozovdd/fal-ai-skill

Gemini CLI

gemini extensions install https://github.com/morozovdd/fal-ai-skill

Prerequisites

• FAL_KEY — get one at fal.ai/dashboard/keys
• bash, curl, jq (standard on macOS and Linux)

What's Inside

Skills

Skill	Purpose
fal	Image/video/audio/3D generation, editing, analysis, model discovery, pricing, training
fal-workflow	Multi-model pipeline JSON authoring with validation rules

Scripts

All scripts live in skills/fal/scripts/ and share skills/fal/lib/common.sh for auth and HTTP helpers.

Script	Purpose
search.sh	Discover models by query and category
schema.sh	Fetch model input/output parameters
pricing.sh	Compare model pricing
run.sh	Run a model synchronously
queue.sh	Submit, check, and manage async jobs
upload.sh	Upload files to fal CDN
docs.sh	Search fal documentation

The Basic Workflow

1. fal skill activates when you mention generating, editing, or analyzing media
2. Agent discovers models, presents options with pricing, waits for your choice
3. Runs a cheap probe, shows you the result
4. You refine (adjust prompt, settings, or try a different model) as many times as needed
5. Final execution at full quality

For multi-model pipelines (e.g., generate image → animate → add audio), the fal-workflow skill handles structured JSON pipeline authoring.

License

MIT