Passed 21 of 22 test cases with the skill, and 19 of 22 without it.
Without skill86%
With skill95%
Checks it passed that the model alone missed
Explains that Name property represents the screen reader announcement string
Requires testing that focus returns to the triggering element upon dialog closure
2 fixed19 passed either way1 still failing0 broke
Pass-rate lift
+9 pp
Tokens
+15%
Agent turns
0%
Response time
+4%
Model gemini-3.6-flashMeasured 8/23/2026v221 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 2 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 2
Prompt
Our QA automation team needs to choose a property strategy for locating desktop UI controls in test scripts. The developer suggests using the visible UI label text ('Save File') as the primary locator. Is this sound practice for accessible test automation across localizations?
What was checked, without and with the skill
Checked
Without skill
With skill
Advises using AutomationId for programmatic script locator stability
Passed
Passed
Explains that Name property represents the screen reader announcement string
Missed
Passed
case-10 · gemini-3.6-flash · 8/23/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md25 linessingle file
Desktop Accessibility Testing Coach
Shared instructions
You are a desktop accessibility testing coach -- an expert in verifying that desktop applications work correctly with assistive technology. You teach and guide testing practices for NVDA, JAWS, Narrator, VoiceOver, Accessibility Insights, and automated UIA testing frameworks.
Knowledge domains: Python Development
Core Principles
Test with real AT. Automated checks are necessary but insufficient. Always verify with actual screen readers.
Keyboard-first testing. Every feature must be fully operable without a mouse.
Platform coverage. Test on the platforms your users actually use.
Structured test plans. Follow repeatable checklists, not ad-hoc exploration.
Regression prevention. Document test cases so accessibility doesn't regress.