Documentation Index
Fetch the complete documentation index at: https://quashbugs.com/docs/llms.txt
Use this file to discover all available pages before exploring further.
The quality of your prompts directly determines the quality of the tests Megumi generates. A vague prompt produces generic tests. A specific, well-structured prompt produces tests you can run immediately.
Megumi generates tests by inferring intent from what you write. The more clearly you express what you want to test, what the expected outcomes are, and which edge cases matter, the less inferring the agent has to do — and the more accurately the output reflects your actual intent.
You do not need to write code. You do not need to know test frameworks. You need to write clearly.
What to include in every prompt
Every prompt in Recipe should include four things:
The feature or flow you are testing. Be explicit. Name it. Do not say “the main screen” — say “the product listing screen”. Do not say “the flow” — say “the checkout flow from cart to order confirmation”.
The scenarios you want covered. List them. Recipe will cover more than you list, but listing the key scenarios ensures they are definitely included. Think: happy path, sad path, edge cases, validation errors.
The expected outcomes. What should happen when each scenario runs correctly? Include this. Expected outcomes become the assertions in the test case — the actual pass/fail criteria. Without them, a test confirms that steps ran but not that they worked correctly.
UI element names, if you know them. If you know the button is called “Sign In” (not “Login”), say so. If the field is labelled “Mobile Number” (not “Phone”), use that. Specificity here makes tests directly executable rather than requiring interpretation.
Prompt length
There is no character limit on prompts. Write as much detail as you need. A longer, well-structured prompt almost always produces better output than a short vague one. If you have a lot of scenarios to cover, write them all out — numbered lists, bullet points, and plain paragraphs all work.
Prompts and CONFIG work together
Your prompt tells Megumi what to test. The CONFIG panel tells it how thoroughly, what type, and on which platform. A prompt saying “generate tests for the checkout flow” produces significantly different output depending on whether CONFIG is set to Minimal + Feature or Full + Integration. Set CONFIG before your first prompt to shape everything that follows.
→ Configuration in Test Studio
In this section
Writing effective prompts — how expected outcomes become assertions, how to use multiple context sources together, how to ask Megumi questions beyond test generation, how to refine tests after the first pass, and how to structure long prompts.
→ Writing effective prompts
Prompt patterns & examples — five proven patterns with templates and worked examples: feature + scenarios, user journey, edge cases, spec-driven, and multi-source generation.
→ Prompt patterns & examples
Common mistakes — the most frequent prompting errors and how to fix them, with a pre-prompt checklist.
→ Common mistakes