model-recommendation

What it does

Analyzes a target .agent.md or .prompt.md file to understand its purpose, complexity, and required capabilities, then recommends the most suitable AI model(s) from GitHub Copilot's options. Provides rationale based on task characteristics, cost efficiency, and performance trade-offs.

How it works

• Reads the target file and extracts frontmatter (description, mode, tools, model if specified).
• Analyzes body content to determine task characteristics: complexity, reasoning depth, code generation needs, multi-turn requirements, context window, and specialized capabilities.
• Categorizes the task into one of eight task types (Simple Repetitive Tasks, Code Generation & Implementation, Complex Refactoring & Architecture, Debugging & Problem-Solving, Planning & Research, Code Review & Quality Analysis, Specialized Domain Tasks, Advanced Reasoning & Multi-Step Workflows).
• Determines capability requirements from tools in frontmatter and body: read-only, write, execution, advanced, multi-modal.
• Evaluates available models against a decision tree considering task complexity, reasoning depth, code needs, context, vision needs, and subscription tier costs.
• Generates a primary model recommendation with rationale and cost implications, plus 1–2 alternative options with trade-offs.
• Checks for auto-selection suitability and notes limitations; flags deprecated models and provides migration guidance and deprecation timelines.
• Considers subscription tier (Free, Pro, Pro+) to frame cost implications and permissible multipliers.
• Advises on frontmatter updates to lock in the recommended model and aligns tool usage with the chosen model.

When to use it

• When you want an automated, model-appropriate recommendation based on the content and intended use of a .agent.md or .prompt.md file.
• When considering model compatibility with tools such as read-only, edit, run commands, or context7/sequential tools, and needing to factor in subscription tier constraints.

What it can touch

• Frontmatter fields including mode, tools, and model (if specified).
• It references the model capabilities matrix and the decision tree to make recommendations, and it may suggest frontmatter edits.
• It assumes access to the listed available models and their multipliers, costs, and capabilities as described in the workflow.

Caveats

• Recommendations are based on the explicit task analysis and the provided decision tree; no outcomes are guaranteed.
• If the file specifies a deprecated model, the guidance flags deprecation and suggests alternatives with migration notes.
• Context7 usage is suggested for uncertain capabilities and requires external lookup; results are only as up-to-date as Context7 data.