Prompt Engineering Coach

Description

Get expert feedback on your AI prompts. Paste an existing prompt, select the target model family (OpenAI, Anthropic, or Google), and receive a scored evaluation with specific improvements and an enhanced version.

Good Query Examples

• "Review this prompt and suggest improvements: [prompt text]" (model: anthropic)
• "Optimize this prompt for GPT-4: [prompt text]" (model: openai)

Bad Query Examples

• "Write me a prompt from scratch" → This agent coaches on existing prompts, it doesn't generate new ones.
• "Run this prompt and show me the results" → This agent evaluates prompts, it doesn't execute them. Use the AI model directly.

Limitations

• Requires an existing prompt to evaluate
• Cannot test or execute prompts
• Model-specific advice limited to OpenAI, Anthropic, and Google families

Input/Output Spec

Input:

• prompt (required): The prompt text to evaluate
• model_family (required): "openai", "anthropic", or "google"
• context (optional): What the prompt is used for
• issues (optional): Known problems or areas of concern

Output:

• Evaluation score with checklist assessment
• Model-specific recommendations
• Enhanced prompt version with reasoning

Disambiguation Signals

• This agent is standalone and should not be included in multi-step workflows. Only activated when the user explicitly asks for prompt evaluation or prompt engineering help.