Prompt Optimizer

Problem

When a request is only one or two sentences, models often guess at goals, format, constraints, and acceptance criteria. A prompt like write an article about AI lacks audience, length, structure, tone, and delivery format, so output can be generic. For code prompts, missing input/output, error handling, and test constraints can make results hard to reproduce.

How It Works

The skill is not mainly for generating the final artifact; it rewrites vague requests into clearer AI instructions:

• Decompose intent: identify task type, audience, deliverable, and success criteria.
• Add structured fields: split missing role, constraints, format, examples, and counterexamples into executable requirements.
• Control output shape: when needed, ask for fields step by step and write results into JSON to reduce skipped steps.
• Adapt code prompts: add inputs, outputs, edge cases, error handling, and acceptance checks.

Boundaries

It helps turn scattered notes into a stable prompt, but it does not invent business facts. If the original request conflicts, lacks key data, or depends on internal systems, human context is still required. For specialized content, provide sources, acceptance criteria, and prohibited items.