Unit coverage
A branch's Unit tab shows which functions your unit tests cover, which they miss, and gives you a prompt that closes each gap.
Read the Unit tab
Open a branch and click Unit. The numbers at the top count the functions in view, and the list below groups them the same way:
- Untested: no test runs any line of the function.
- Partially tested: tests run some of it, but lines or branches are missed.
- Fully tested: tests run all of it.
- Failing: tests that failed in the last run. This count and the Failing tests section appear only when there are some.
To narrow the list:
- Changed only shows the functions this branch added or modified. All functions shows the whole codebase. The switch appears when the branch can be compared with your default branch.
- Tree groups files by folder. Flat lists every file by its full path.
- The filter box searches by function or file name.
Look at one function
Click a function to open its page. It shows:
- The function's source, with the lines no test reached marked.
- Show diff, for a function the branch modified: the change against the default branch.
- Tests: the tests that reach the function, and whether each passed.
- A switch between Unit Tests, Browser Coverage and Manual Recording, when the function has data for more than one. The last two are covered in Code reached.
Get a prompt that closes the gap
A prompt is an instruction you paste into your coding agent. There are two kinds, both on the page of a function that is not fully tested:
- Baseline prompt: free and instant. It is built from the function and its uncovered lines, so it is the same every time you ask.
- Generate LLM prompt to fix issuesPaid plan: written by an AI model that reads the function, your existing tests and the test conventions saved in the repository's settings.
To handle many gaps at once, use the bulk prompts on the Unit tabPaid plan:
- Generate prompt for everything wrong: every failing test and every untested or partially tested function.
- Generate prompt for frontend gaps: every React component that still has an untested handler or render branch.
- The button on the Failing tests, Untested and Partially tested headings writes one prompt for that section.
Your agent gets the same prompts over MCP, without you copying anything: get_baseline_prompt is the free one, and get_generated_prompt, get_bulk_prompt and fix_branch are the paid ones. See MCP tools.
Coverage reports verifAIed accepts
- Python: pytest-cov's
coverage.json. - JavaScript and TypeScript: Istanbul's
coverage-final.json, which is what vitest's Istanbul provider writes. The v8 provider is not accepted.
Other formats, such as LCOV and Cobertura XML, are rejected. For Python, run the tests like this so the report says which test ran which line:
pytest --cov --cov-branch --cov-context=test --junitxml=junit.xml
coverage json --show-contexts -o coverage.json
The second command matters. Without it the report has no per-test data, and the Tests list on a function stays empty.
Failing tests are known only when you send the JUnit file too. From your machine, pass it with --junit:
verifaied upload --junit junit.xml
CI coverage shows how to upload from GitHub Actions, and CLI commands lists every option of verifaied upload.