UTILITYPILOT · THE DETAILS
How we test the tools
Last reviewed: 17 November 2026.
What “tested” means here
“Tested” describes specific checks that were run, not a guarantee that every input works. This page lists what is checked, how, and what remains unverified so you can judge a result for your own workflow.
Automated processor tests
Each local text, data, image, web and developer processor has unit tests covering expected results and boundary cases: empty input, maximum sizes, malformed records, Unicode boundaries and failure paths. The suite runs before a release build is accepted. A passing suite shows that covered cases behave as specified; it does not cover every possible document.
Browser and interface checks
A scripted browser suite opens every published route, runs each tool, and checks canonical URLs, one main heading, copy, download and reset behaviour, console errors, unexpected external requests and layout at 320, 390 and 768 CSS pixels. Checks are run in Chromium-based browsers and Microsoft Edge.
Accessibility checks
Automated accessibility rules (WCAG 2 A/AA and 2.1 AA) run across the site, and focus behaviour is checked for keyboard reordering controls. Automated rules find some defects, not all. Full screen-reader testing with real users and a complete assistive-technology matrix have not been performed.
Category-specific checks
- Text and lists: grapheme and code-point counting, CRLF and LF line endings, combining marks, empty and maximum-length input.
- Data and tables: quoted delimiters, embedded newlines, duplicate or blank headers, differing record widths, large identifiers and typed-value boundaries.
- Images: rotated photographs, transparent PNGs, JPEGs without alpha, impossible byte targets and actual exported format and size.
- Web and colour: malformed URLs, encoded query values, mixed-case input and contrast pairs at the WCAG AA boundary.
- Developer tools: integers beyond JavaScript’s safe range, invalid or non-compact tokens, timezone offsets and regular expressions stopped by a worker timeout.
What is not tested
- Acceptance of an export by every external application.
- Every browser version, device and operating system.
- Correctness of the underlying material you paste or select.
- Professional, legal, security, financial or accessibility certification.
Reproducing and reporting
Use invented data, note the tool and the exact steps, and include the expected and observed result. The report helper builds a reproducible structure locally. Confirmed defects are fixed and recorded on the changelog.