Loading page
About this tool
Builds a test-generation prompt with your framework, coverage target and edge-case classes, plus a baseline test-count estimate.
The Unit Test Prompt Builder generates a test-writing prompt tailored to your language, test framework, branch-coverage target and chosen edge-case classes — empty input, boundary values, error paths, unicode, async ordering and more, following classic boundary-value analysis. It estimates a baseline test count (one happy path plus one test per edge-case class per public function) and writes framework-specific conventions — pytest fixtures, Jest matchers, Go table-driven tests, JUnit 5 parameterised tests — directly into the prompt.
Open Unit Test Prompt Builder on AltFTool — it loads instantly in your browser.
Paste or upload your source content.
Choose the target format or conversion mode.
Copy or download the converted result.
The prompt embeds each framework's own conventions — test file naming, assertion style, parameterisation mechanism — so generated tests drop into your suite unchanged.
Eight selectable equivalence classes from boundary-value analysis turn 'add some edge cases' into a specific, checkable list.
Generated tests must assert specific values and error types, never 'does not throw', and must flag suspected bugs rather than enshrine them.
A practical baseline is one happy-path test plus one test per relevant edge-case class — so a function checked for empty input, boundary values and error paths gets four tests. This tool multiplies that per-function figure by your function count; parameterised or table-driven tests can then merge similar cases without losing coverage.
Google's published code-coverage guidance describes 60% as low, 75% as acceptable and 90% as exemplary, measured meaningfully as branch coverage rather than line coverage. Chasing 100% usually produces assertion-free tests; this tool's prompt instead asks the model to list branches it could not cover from the public API and why.
Constrain the prompt: name the framework and its idioms, enumerate the edge-case classes to cover, require Arrange-Act-Assert structure with one behaviour per test, and ban weak assertions like 'is truthy'. This tool generates exactly those constraints, plus a rule that tests must not depend on time, network or execution order.
No. The generated prompt instructs the model to mock only what crosses a process or network boundary — databases, HTTP APIs, the filesystem — and to use real objects everywhere else, because over-mocked tests verify the mocks rather than the code.