--- title: Example prompts description: >- Explore ready-to-use Kapptivate prompts for investigating failures, creating tests, running campaigns, and tracking quality over time. sidebarTitle: Examples lastUpdated: "2026-09-23" --- > **For AI agents:** the complete documentation index is at [llms.txt](/llms.txt). Append `.md` to any page URL for its markdown version. Once an assistant is connected, you drive Kapptivate in plain language. These are prompts that work well, grouped by goal. You rarely name a tool: describe the outcome and the agent picks the path. For anything that creates or changes data, the agent shows you the plan first. ## Investigate a failure The fastest way to understand *why* something broke, when the dashboard only tells you *that* it broke. - "Monitor 4521 went red an hour ago. What changed, and what is the most likely cause?" - "Summarize the failures in last night's checkout regression, grouped by root cause." - "Open execution 889123, find the failed step, and show me its error and screenshot." - "Which monitors are in an open incident right now, and what do they have in common?" - "Compare the last passing run of test 16480 with the first failing one and tell me what differs." ## Create a test Describe the test in plain language and the agent scaffolds it (it reads a template, fills it, and creates the test on review). Works for every channel. - **API:** "Create an API test on acme~web that sends GET https://api.acme.test/health and asserts the status is 200." - **Web:** "Create a web test on acme~web that opens the login page, signs in as the demo user, and checks the dashboard heading is visible." - **Smartphone app:** "Create a smartphone app test that launches our Android app, logs in, and verifies the account balance screen appears." - **USSD:** "Create a USSD test that dials *123# and waits for the balance SMS to arrive." - **SMS:** "Create an SMS test that sends BAL to 1234 and checks the reply contains the balance." Be specific about the product slug, the endpoint or screen, and what counts as success. The agent maps it to the right action type (API, web, app, USSD/SMS) and asks before it writes. ## Run a test or campaign - "Run the checkout regression on Staging and tell me which tests failed." - "Build a campaign from tests 16480 and 16838 on acme~web, pin the API group to Prod, run it, and wait for the results." ## Track quality over time - "What is the p95 response time trend for the payments product this week?" - "Are we on track for our 99.9% SLO this month, and how much error budget is left?" ## Manage configuration - "Add a Staging variant to the API variable group and set BASE_URL to https://staging.acme.test." For a multi-step task, a good opening prompt is "load the platform guide for test creation before you start" — the server exposes a `platform_guide` tool that returns the exact, current workflow. ## What's next? Set up Claude, Cursor, or Claude Code. Operators, products, variables, and the safety model. The full tool catalog behind these prompts.