## Usage Guide

1. **Prepare your context** – Fill in the placeholders (`{{project_name}}`, `{{problem_statement}}`, etc.) in the prompt body.
2. **Select a model** – Use any of the supported models; they all understand the `role:` syntax.
3. **Run the prompt** – Paste the entire block into your LLM UI (ChatGPT, Claude, Gemini) and hit generate.
4. **Validate output** – The response should be a single JSON object. Paste it into a JSON validator to ensure correctness.
5. **Iterate** – If the `research` section returns a `needs_more_info` flag, provide the missing details and re‑run.
