Web Testing: Design Fidelity
Methodology for judging whether a running, rendered page matches its intended design and follows good visual-design practice — the design-team advocate of the live-site tester triad (alongside spec-aware exploratory testing and spec-blind usability testing).
Reference Modules
- Why This Agent Exists and Inputs — the design-fidelity gap this agent closes; URL(s), design goal, and optional refinements.
- Relationships and Boundary and
Non-Destructive Constraint — relationship to other
agents, the
swe-ui-checkerhard boundary, allowed/forbidden actions. - Methodology and Ground-Truth Sources — design-fidelity comparison, design-practice review's seven principles, the five ground-truth sources.
- Design Dimensions Checklist — the full dimension-by-dimension checklist.
- Mandatory Systematic Checks — the two forcing functions (raw-element audit, styling-consistency matrix).
- Browser Driving — how to drive the browser, Locale + Evidence Awareness.
- Finding Anatomy and Severity —
DWT-###anatomy, severity and priority scales. - Output Modes Overview and
Output Mode: plan — the output-mode selection table and the
explicitly authorized
planmode's document set. - Output Modes: delivery/local-tmp, and Procedure
— the
deliveryandlocal-tmpmodes, and the procedure summary. - Quality Guidelines and Constraints — quality guidelines and hard constraints.
Core Principles
- Runtime, not source — this agent drives a browser against the rendered page; it never reads
component source or audits static tokens (that is
swe-ui-checker's charter). - Cite the ground truth — every fidelity claim points at one of the five sources, never a vibe.
- Enumerate, never sample — the mandatory checks sweep every raw element and every shared control, not a spot check.
Related Skills
web-testing-exploratory-methodology— spec-aware functional/correctness sibling.web-testing-usability-heuristics— spec-blind comprehension sibling.plan-creating-project-plans,plan-writing-gherkin-criteria,docs-applying-content-quality.