Prompt Engineering Assistant

Problem

Prompt writing often fails because outputs drift: models ignore field names, skip reasoning steps, blur role and task boundaries, or lose constraints across turns. This skill turns prompt design into reusable templates, helping you converge on input structure before iterating on examples.

How It Works

It acts as a prompt engineering expert and produces practical prompt proposals from your task goal, input-output examples, and constraints. Core capabilities include:
- System Prompt templates: separate role, task, constraints, and output format to reduce conflicting instructions.
- Few-shot learning: use small examples to stabilize style and boundaries for classification, extraction, or rewriting.
- Chain-of-thought: ask the model to reason first, then answer, which can reduce missed conditions in complex judgments.
- Structured output: constrain JSON, field names, enums, and other parseable formats for downstream code.
- Role-based prompting: define tone, expertise, and scope through a role.

Use it by describing the business scenario and expected output, then ask it to draft or revise the prompt. It provides prompt engineering methods, not model capability; for private rules, compliance constraints, or high-consistency outputs, include acceptance examples, negative cases, and error handling, and validate with real-world regression tests.