Developer Tools

Prompt Engineering Playground

Opinionated prompts I actually use for senior engineering work — code review, ADRs, postmortems, Terraform, Kubernetes, RCA, system design. Fill in the variables, copy the rendered prompt.

Free Tool100% Local12 templates
Privacy: templates render in your browser. Your variable inputs never leave the page.

Templates

Click a template above to open the playground.

Why structured prompts beat free-form

The same model answers the same question very differently depending on how the prompt is framed. After running these in production for a while, three patterns emerged that are baked into the templates here:

  • Specify the output shape. "Give me a postmortem" is bad. "Give me a postmortem with these exact sections in this order" is good. The model uses the structure as a checklist.
  • Ask for what NOT to do as well. Constraints reduce hallucination more than instructions add it.
  • Reusable variables. The interesting work is filling in project-specific context, not rewriting the scaffolding every time.

How to use

  1. Pick a template.
  2. Fill in the variables.
  3. Copy the rendered prompt into Claude, ChatGPT or your IDE's AI panel.
  4. Iterate on the variable values, not the template. If the template needs changes, fork it into your own playbook (Notion, Obsidian, repo).

Frequently Asked Questions

Are these prompts model-specific?
Written to work across Claude, GPT and Gemini frontier models. Each template lists which I have actually tested it with.
Why opinionated structure instead of free-form?
Structure reduces output variance. When every PR description follows the same shape, reviewers cognitively cache the format and focus on substance.
Can I save my customised prompts?
Not yet. The tool is intentionally stateless — copy the rendered prompt to your own playbook.
Does this tool send my inputs anywhere?
No. Templates are bundled with the page; variable substitution happens in your browser. Open DevTools → Network — no requests fire.