I will do ai prompt engineering, prompt writing and system prompt

Most folks have a specification problem - they tell the model what to write instead of what "correct" looks like. So it works then drifts and can't tell why.

I write the spec first, then test it until it holds.

WHAT YOU GET

- The prompt, copy-paste ready, with variables marked so your team can reuse it

- A system prompt if you're building an assistant, custom GPT or automation

- Usage notes: which model, what to change, what good output looks like

- The edge cases I tested it against, and what it did

- Free retuning if it drifts within 14 days

WORKS WITH

ChatGPT (GPT-4 / GPT-5), Claude, Gemini, Grok, Llama - and prompts built for n8n, Make, custom GPTs or in-app assistants.

TYPICAL JOBS

Support replies that stay on policy. Content that stays on brand. Data extraction that returns the same format every time. Agent instructions that stop looping.

Send me one output you'd call correct and one you'd call wrong. That's usually all I need to start.