Skip to main content
Download: How to Prompt Guide (May 2026) (PDF)
Browse all prompting guides
The difference matters. Recipe prompts are about describing what to test. Task prompts are about describing what to do. The more precisely you write a task prompt, the more consistently Mahoraga executes it.

Start with shared fundamentals

Apply the shared fundamentals from Overview first: name exact UI targets, define expected outcomes, handle interruptions, and keep scope focused.

The structure that works

Every good task prompt follows the same shape:
Start with where to go. Then what to do. Then what should be true at the end. This gives Mahoraga a clear sequence with a defined success condition.

Examples

Simple verification:
Data modification:
Multi-step flow:
Negative test:
First-run flow with interruptions:

Execution-specific guidance

1. Use credentials from Apps

If credentials are configured in Apps → Credentials, reference them by name instead of hardcoding login details:
Mahoraga pulls the stored credentials automatically. This keeps prompts clean and means you never need to update prompts when passwords change — just update the credential set in Apps.

2. Be explicit about navigation

Do not assume Mahoraga knows the fastest path between screens. Describe the navigation steps if the route matters for the test:
If the route does not matter and you just need to arrive at a screen, a simpler instruction works:
Mahoraga will find its own path. Use explicit navigation when the route itself is what you are testing (e.g., “verify the back button returns to the previous screen, not the home screen”).

Using test data with /slug

If the same task needs to run with different inputs — different user accounts, different search terms, different product selections — use Test Data instead of writing separate tasks. Type / in the task prompt input. A dropdown appears listing available datasets with their slugs and sizes. Select one and the slug is inserted at the cursor. When the task runs, Mahoraga executes it once per row in the dataset automatically. Example:
If /login-credentials has five rows, the task runs five times — once per set of credentials. → Test Data For failure patterns and corrective rewrites, see Troubleshooting Prompts.

When to use tasks vs. Test Studio

Tasks are ephemeral — run once, read the report, move on. If you find yourself running the same task repeatedly, save it as a test case and add it to a suite. → Test Cases → Suites