Week 45, updated 2 Nov 2026

AI use cases Coding agents AI test writing

AI test writing

Generate unit and integration tests with a coding agent, then run them and keep only the ones that mean something. Prompts, workflow and ranked tools.

Tools compared
4
Open source
3of 4
Reviewed
16 Oct

For: Developers

Best-scored tools for this jobscore, safety
  1. 1Claude CodeAI tool65No dependencies scanned
  2. 2ClineAI tool65Caution
  3. 3ContinueAI tool37Caution
  4. 4Roo CodeAI tool23Caution
What the score and shapes mean

Score = public maintenance, adoption and docs signals, as a percentile of 52 tools. Nobody here tested these hands-on. Shapes are the separate safety label: No issues found, No dependencies scanned, Caution.

What it does for you

Developers with untested or under-tested code, and anyone fixing a bug who wants a test to prove it.

  • Draft unit tests for existing functions
  • Add regression tests that reproduce a reported bug
  • Suggest edge cases you had not listed
  • Fill in test fixtures and mocks
  • Run the suite and fix flaky or failing tests
More about this job

Agents can draft tests quickly and run them in the same loop, which makes them useful for adding coverage to untested code. The value is in the review: a test that always passes proves nothing.

Workflow in 5 steps

  1. Pick one module and say which behaviour matters.
  2. Ask for a failing test first when fixing a bug, then the fix.
  3. Run the new tests.
  4. Delete tests that only repeat the implementation or assert nothing.
  5. Commit tests separately from code changes.
Full steps
  1. Pick one module and say which behaviour matters.
  2. Ask for a failing test first when fixing a bug, then the fix.
  3. Run the new tests; break the code on purpose to confirm they can fail.
  4. Delete tests that only repeat the implementation or assert nothing.
  5. Commit tests separately from code changes.

Prompts to copy

Starting points we wrote, not tested results.

Write unit tests for parse_invoice() in billing/parser.py. Cover empty input, malformed dates and negative amounts. Run them and report which fail.

2 more prompts

Reproduce this bug as a failing test first: users with a trailing space in their email cannot log in. Do not fix the bug yet.

Review the tests in tests/test_cart.py and list any that would still pass if the function under test were deleted.

Watch out for

  • Treat AI changes like a pull request from a new colleague: read the diff, run the tests and keep a human sign-off before anything ships.
  • AI output can be fluent and wrong.
  • Generated tests can encode the current (possibly wrong) behaviour as correct.
  • Licence to check before commercial use: Claude Code (Proprietary (Anthropic Commercial Terms)).
  • No dependencies scannedClaude CodeCode checks clean, no lockfile to check
  • CautionCline13 critical-severity known vulnerabilities in runtime dependencies (OSV)
  • CautionContinue36 critical-severity known vulnerabilities in runtime dependencies (OSV)
  • CautionRoo Code13 critical-severity known vulnerabilities in runtime dependencies (OSV)
Why these warnings?
  • Treat AI changes like a pull request from a new colleague: read the diff, run the tests and keep a human sign-off before anything ships.
  • AI output can be fluent and wrong. Check facts, code and citations yourself before you rely on them or publish them.
  • Generated tests can encode the current (possibly wrong) behaviour as correct. Decide the expected behaviour yourself.
  • Claude Code: code and instruction-file checks found nothing, but there is no runtime lockfile or manifest, so dependencies were not checked
  • Cline: 13 critical-severity known vulnerabilities in runtime dependencies (OSV)
  • Continue: 36 critical-severity known vulnerabilities in runtime dependencies (OSV)
  • Roo Code: 13 critical-severity known vulnerabilities in runtime dependencies (OSV); Repository archived

Recommended tools

From our ranked data, highest score first. A high score does not mean a clean scan.

  • 65Claude CodeAI tool · Proprietary (Anthropic Commercial Terms) · #11 overall
  • 65ClineAI tool · Apache-2.0 · #12 overall
  • 37ContinueAI tool · Apache-2.0 · #43 overall
  • 23Roo CodeAI tool · Apache-2.0 · #51 overall

Related jobs

Sources for tool facts

anthropics/claude-code · cline/cline · continuedev/continue · RooCodeInc/Roo-Code. Data fetched 2 Nov 2026. Scores and labels are automated, relative and not a hands-on test.

Get the board every Monday

One email a week: the new top 10, the biggest score changes and every safety label that changed.

Free. Double opt-in: you confirm by email first, unsubscribe in one click.