[SKILL]: hf-create-a-space for deploying models to HF Spaces

[GhostScientist]

This PR adds a new skill for creating and deploying Hugging Face Spaces. The skill guides users through choosing the correct deployment method and warns about common compatibility issues before they occur.

This PR includes various templates to help build various app for the space depending on your model's use case.

Key Features

• Deployment method selection upfront - Asks users to choose between:
  • Inference API
  • ZeroGPU
• Compatibility checks - Validates before deployment:
  • Is model from a supported provider? (Qwen, meta-llama, mistralai, etc.)
  • Is it a LoRA adapter? (forces ZeroGPU, identifies base model)
  • Is it a gated model? (warns about HF_TOKEN requirement/potential need to accept conditions)
  • Model size vs hardware limits
• Post-deployment management - Scripts for:
  • Changing hardware runtime (manage_space.py hardware)
  • Adding secrets (manage_space.py secret)
  • Monitoring build status (monitor_space.py)
  • Auto-remediation of common errors (remediate.py)

File	Purpose
SKILL.md	Skill instructions, templates, troubleshooting guide
scripts/deploy_model.py	Main deployment with --method inference-api
scripts/manage_space.py	Change hardware, add secrets, pause/restart
scripts/monitor_space.py	Build monitoring & error detection
scripts/preflight.py	Pre-deployment validation (token, model access)
scripts/remediate.py	Auto-fix common deployment issues
templates/	Gradio app templates (chat, LoRA, ZeroGPU)

Open to any and all feedback.

[GhostScientist]

I used this skill to deploy a fine-tuning to a Chat UI space.

https://huggingface.co/spaces/GhostScientist/qwen-coder-assistant

[gary149]

interesting, it should be more focused on how to connect to ssh/dev mode/get the logs/zerogpu imo so the agent has the complete flow (name should just be hf-spaces probably)

https://huggingface.co/spaces/GhostScientist/qwen-coder-assistant

nice!