The Prompt Field Guide
A prompt is the written instruction that tells an AI system how to act. Good prompts make AI answers clear and safe. Weak prompts cause vague or risky replies.
This free tool teaches you to spot the difference. Paste a practice prompt below. In about a minute you get a score and concrete fixes, based on the rubric ICX uses in client work. No sign-up needed.
New · Built on Google Cloud
See what makes a prompt work.
This learning lab shows what makes a prompt easy to understand and test. It is a field guide, not a grade from the prompt police.
Your field notes will appear here
Five useful questions. No magic formula.
- Is the task clear?
- Does the model have the right context?
- Are boundaries visible?
- Is the output shape defined?
- Could you tell whether it worked?
Reading the field notes
Gemini is checking the prompt against a teaching rubric…
Heuristic learning audit
Field notes
This score is a teaching aid, not a certificate. It does not predict how a model will perform.
Three moves to try next
Teaching example
An improved version to study
Anthropic recommends clear, direct instructions, examples, and visible structure. XML tags help separate instructions, context, and examples in complex prompts.
Read Anthropic’s official guidance ↗OpenAI recommends clear instructions, useful context, and a stated output format, plus examples where they help.
Read OpenAI’s official guidance ↗Google treats prompt design as a test-driven loop: set the goal, structure the input, add safeguards, and check the results.
Read Google’s official guidance ↗Practice and education only. Do not use this result as approval for live, confidential, regulated, or production work.