Web QA thatshows its work.
A small workflow, a deliberate positive-control defect, and the exact evidence needed to reproduce, fix, and retest it.
Assertion WEB-007
Not runPreserve the selected role after inline validation fails.
Expected
Role remains Admin
Actual
Run the validation path
The positive-control build intentionally resets Admin to Viewer. The fixed reference preserves state. This proves the check can fail before a clean run is trusted.
Execution summary
Calibrate, exercise, reproduce, retest.
Downloadable evidence
Inspect the plan, result, defect, and runner.
The artifacts use synthetic data and carry their own scope and limitations. The JavaScript harness runs without a test framework so the result is easy to reproduce.
Requirement-to-test matrix
Twelve checks. One intentional failure.
The same assertions run against both builds. Only WEB-007 changes outcome, which localizes the regression and prevents a vague “looks fixed” conclusion.
| ID | Assertion | Control | Seeded | Fixed |
|---|---|---|---|---|
| WEB-001 | Valid invite | Happy path | PASS | PASS |
| WEB-002 | Required name | Negative | PASS | PASS |
| WEB-003 | Invalid email partitions | Negative | PASS | PASS |
| WEB-004 | Trim + normalize email | Data integrity | PASS | PASS |
| WEB-005 | Duplicate email | Business rule | PASS | PASS |
| WEB-006 | Unknown role | Authorization | PASS | PASS |
| WEB-007 | Role survives validation error | State | FAIL | PASS |
| WEB-008 | 80-character name | Boundary | PASS | PASS |
| WEB-009 | 81-character name | Boundary | PASS | PASS |
| WEB-010 | Render untrusted text inertly | Safety | PASS | PASS |
| WEB-011 | Idempotent double submit | Reliability | PASS | PASS |
| WEB-012 | Accepted role in result | Regression | PASS | PASS |
Defect QA-001
State silently falls back after validation.
Severity: medium. A user who selected Admin can unknowingly submit a lower-privilege Viewer invite after correcting an invalid email.
Honest limits
Evidence is useful only when its boundary is visible.
Need a workflow tested this way?
Define the environment and acceptance boundary. Receive cases, evidence, defects, regression results, and explicit exclusions.