home/writing/prompt-design-11

Prompt Design

GPTClaudeGemini··▲ 1,342 copies·updated 2026-07-14

prompt-design-11.prompt

# Prompt Design

Each task in this project is backed by an explicit prompt template and schema.

## Components

1. **Base Prompt Template** (Markdown)
   - Describes the task (classification, summarization, extraction).
   - Specifies the required JSON output structure.
   - Includes guardrail instructions (no speculation, no extra commentary, etc.).

2. **Few-shot Examples** (JSON)
   - Simple `(input, output)` pairs stored in `prompts/few_shot_examples/*.json`.
   - Can be embedded into prompts or used to sanity check behavior.

3. **Pydantic Schema**
   - Defines the structured output expected from the LLM.
   - Parsed and validated at evaluation time.
   - Allows us to treat malformed outputs as explicit failures.

## Goals

- Make prompt contracts **explicit and enforceable**.
- Encourage **structured outputs** over free-form text.
- Provide a stable foundation for **prompt iteration and regression testing**.

when to use it

Community prompt sourced from the open-source GitHub repo moses-shenassa/llm-prompt-framework-and-eval-suite (MIT). A "Prompt Design" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.