- Industry
- Procure AI
- Department
- Engineering
- Team
- Application Engineering
- Kind
- Assistant
- Approver
- QA Lead
- Code
- ENG-03
It writes automated unit, integration and end-to-end tests for platform changes and the core journeys (requisition to approval, PO to goods receipt, invoice to three-way match, sourcing event to award, supplier onboarding to screening), runs them in CI and reports failures, flaky tests and untested behaviour. It quarantines a flaky test only within the policy and never weakens an assertion to make a build pass.
How it works
6 steps, and a person signs off before step 6.
- 01Looks it up
Finds the record the request is about in the CI system: the latest test runs, failure and rerun history and the coverage report for the branch named..
- 02Reads the request
Works out what is being asked and what a complete answer needs.
- 03Checks the facts
Checks figures, dates and rules against the source records, not from memory.
- 04Works it through
Prepares the draft, match, recommendation or reply, with its sources.
- 05Asks a person
Hands the work to QA Lead with its reasoning before anything goes out.
- 06Acts
Once approved, it acts in the code repository: it opens a pull request with the new tests and posts the test report on the change. and records what was done.
What it works with
Your knowledge, your systems, your people.
Knows
The test strategy: the test pyramid, coverage targets, the synthetic tenant and supplier data set, the control tests that prove approval limits and segregation of duties, the flaky-test policy and the critical journeys.
Reads from
the CI system: the latest test runs, failure and rerun history and the coverage report for the branch named.
Acts in
the code repository: it opens a pull request with the new tests and posts the test report on the change.
Hands back
An object with tests_added (a list, each with name, level and covers), control_tests (a list of controls proven), run (an object with passed, failed, skipped and flaky counts), failures (a list, each with test, error and likely_cause), coverage_percent (a number or null), gaps (a list) and quarantine (a list, each with test, reason and until as YYYY-MM-DD).
Asks a person when
It proposes quarantining a test in approval routing, three-way match, payment file generation, supplier bank-detail change, tenant isolation, sanctions screening, keeping a test quarantined beyond 14 days, lowering coverage below 80 percent, or using production data.
Settings
Your limits, set by your managers.
| Setting | Default | What it controls |
|---|---|---|
| Coverage target (%) | 80 | Coverage of the code in scope it aims for; a change that lowers it below this is flagged. |
| Reruns to call a test flaky | 2 | A test that fails then passes within this many reruns on the same commit is flaky. |
| Longest quarantine (days) | 14 | A quarantined test is fixed or restored within this long. |
| Journeys that are never quarantined | approval routing, three-way match, payment file generation, supplier bank-detail change, tenant isolation, sanctions screening | Comma-separated. A failing test in any of these blocks the release. |
| Tests may use production data | Off | When off, tests use only the synthetic tenant, supplier and price data set. |
Limits are settings, not code: a manager changes them without a developer.
Rules it always follows
- Write tests at the lowest level that proves the behaviour, one per acceptance criterion, following the test pyramid.
- Every approval-limit, match-tolerance and segregation-of-duties rule has a control test that proves the forbidden path fails: a requester approving their own requisition, an invoice above tolerance auto-matching, a bank change without verification.
- Never delete a test or weaken an assertion to make a build pass; a real failure goes to the change's author.
- Mark a test flaky only when it fails then passes within 2 reruns on the same commit, and quarantine it for at most 14 days.
- A test in approval routing, three-way match, payment file generation, supplier bank-detail change, tenant isolation, sanctions screening is never quarantined; its failure blocks the release.
- Report coverage against 80 percent and list untested behaviour.
- Use production data only when off is true; otherwise use the synthetic data set.
See the Testing Agent agent on your own work.
We’ll run it on a sample of your work and show you what your approver would see.