selecting-models

Resolves installed ComfyUI models for a requested capability via the model registry.

Updated Apr 8, 2026
One-click install
npx skills add https://github.com/ShinyGua/ComfyUI-Agent --skill selecting-models-shinygua
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: selecting-models
Source: https://github.com/ShinyGua/ComfyUI-Agent/tree/main/.claude/skills/selecting-models
Command: npx skills add https://github.com/ShinyGua/ComfyUI-Agent --skill selecting-models-shinygua

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? During workflow planning, the agent needs to know which models are actually installed on the local ComfyUI instance for a given capability, and what to do when none match. This Skill queries the live model registry and returns either a matching ModelHandle or a structured MissingModelEnvelope with sanctioned download candidates. ## Core Features & Use Cases - Capability-based model discovery: Wraps host.bridge.model_registry_client so the planner can ask which models are installed for a given capability before any comfyui_execute sub-goal runs. - Missing model handling: When no model matches, emits a MissingModelEnvelope containing approved download candidates from policies/model_sources.yaml instead of silently substituting. - Use Case: Before executing an image generation workflow, the planner calls this Skill with a capability like checkpoint loading; it receives a ModelHandle (capability, folder, name, matched_glob) or an envelope prompting a model_download sub-goal decision. ## Quick Start Ask the planner which installed ComfyUI models match a given capability before executing a workflow sub-goal.

Frequently Asked Questions about selecting-models

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I check which ComfyUI models are installed for a capability?▼

Run the select script with a JSON payload containing the capability and optional requested_for field over stdin. It builds a ModelRegistry against the live ComfyUI base URL and returns a ModelHandle JSON with capability, folder, name, and matched_glob.

What happens when no ComfyUI model matches a capability?▼

The script emits a MissingModelEnvelope JSON containing sanctioned download candidates from policies/model_sources.yaml. The planner then decides whether to insert a model_download sub-goal or surface the envelope to the user.

Does model resolution require a running ComfyUI instance?▼

Yes, the ModelRegistry is built against the live ComfyUI base URL through the ComfyUIClient bridge. Discovery is registry-driven against the actual backend rather than a hardcoded model list.

What input format does the model selection script expect?▼

The script reads a JSON object from stdin with a required capability string and an optional requested_for string. It exits with an error if capability is missing, and writes the result JSON to stdout.

Can the skill automatically download missing models?▼

No, it never downloads or substitutes models on its own. It only surfaces a MissingModelEnvelope with approved candidates, leaving the download decision to the planner or user.