Runtime UX QA

SkillWeb & browsing

Use when QA-reviewing plans, implementations, screenshots, Playwright specs, or runtime MUI application flows for real user usability. Checks whether a normal user can complete the workflow without raw IDs, raw JSON/object cells, single-line long text, untranslated/internal validation, page-level overflow, inaccessible controls, or hidden product knowledge.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Runtime UX QA skill

What this skill tells your AI

The instructions your AI receives, as published by teknokomo/universo-platformo-react in .agents/skills/runtime-ux-qa/SKILL.md and read by ahel’s review.

Use this skill for QA of UI plans, implementation diffs, screenshots, Playwright browser evidence, runtime CRUD flows, metahub template UI metadata, and published app screens.

Verdict Format

Return:

  • verdict: pass, pass-with-minor-issues, or fail;
  • blockers;
  • majorIssues;
  • minorIssues;
  • passedChecks;
  • browserEvidence;
  • missingEvidence;
  • requiredFixes.

Blocking Fail Criteria

  • A normal user must type or understand a raw user/owner/reference ID.
  • A normal table/card shows raw JSON, [object Object], UUID-only business labels, or internal field names.
  • Semantic long text is single-line.
  • Empty optional resource-source fields show errors.
  • A localized surface shows raw English/internal validation messages.
  • Page-level horizontal overflow appears at realistic desktop/tablet/mobile widths.
  • The workflow cannot be completed by keyboard or user-facing locators.
  • QA only says "tests pass" without answering whether a normal user can use the feature.

Evidence Requirements

  • Prefer browser evidence for implemented UI: screenshots, visible locators, keyboard path, and responsive viewport proof.
  • Automated checks are useful only if they encode UX semantics; CRUD success alone is not enough.
  • When an implementation adds a screen next to an existing canonical screen, compare the shared component contract directly: density, header styling, numbered rows, action placement, and dialog affordances must match unless a visible product reason says otherwise.
  • For destructive confirmation dialogs, browser or component evidence must prove that resize/fullscreen controls are absent and that footer actions have normal spacing. Saving screenshots without assertions is not sufficient evidence for this class of defect.
  • Do not weaken or skip a UX canary without evidence that the failure is environmental.

References

  • Read references/playwright-ux-oracles.md when adding or reviewing browser assertions.
  • Read references/qa-verdict-template.md when producing a structured QA result.
  • Read references/evaluation.md when checking that this skill catches known defect classes.

Signals

GitHub stars
22
Forks
3
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
runtime-ux-qa
Source
github.com/teknokomo/universo-platformo-react