Diagnose visual design issues

Scope

Use visual evidence first, not taste labels. This skill explains why a composition does not convey the intended message. If an observed issue is primarily interaction friction route to ux-diagnose; if it is an accessibility barrier route to wcag-diagnose too.

Procedure

1. Record the surface and audience, user task, design intent, brand constraints, viewport/theme, and observations. Do not infer click behavior, responsive failures or screen-reader behavior from one image.
2. Load references/issue_taxonomy.json: select one primary of ten practical repair categories and zero to two secondary ones. Use references/principles.json to associate at most three of the original seven composition principles. references/elements.json helps describe the material causing the issue (color, texture, space, etc.).
3. Distinguish symptoms from mechanism: for example, "CTA disappears against photo" -> hierarchy/contrast or image treatment, not automatically a color-theory problem. Compare alternatives using the appropriate article summary in references/articles.json and Gestalt concepts where relevant.
4. Read references/cross_principle_conflicts.json for at least one relevant tradeoff. Filter references/remediation_recipes.json to matching category/observation. Load only the matching design-* specialist's references/playbook.md when implementing.
5. Propose the smallest coherent repair that preserves the underlying user task. Test at relevant widths/themes, through grayscale, with real copy and content length. Never claim a design "fails" a subjective principle as if it were a technical standard.

Evidence grades and prioritization

• Confirmed: screenshot or browser/code observation directly supports the issue; observational limits explicitly stated.
• Hypothesis: design mechanism inferred; discriminate with a screen comparison, brief check, measured contrast or user test.
• Prioritize a blocked task or inaccessible content over a weak aesthetic; after those, prioritize task-relevant salience, reproducible confusion and cross-screen visual drift over personal preference.

Output

finding ID | observed symptom | evidence/uncertainty | primary category | relevant design principle(s) | user/brand impact | smallest repair | competing tradeoff | acceptance test. For machine-readable records use references/finding.schema.json.

Boundaries

Digital Polo articles provide design heuristics, not normative accessibility standards. User goals, truthful interface state, consent and applicable WCAG criteria outrank aesthetic optimization. If appropriate, activate ux-* and wcag-* alongside this skill. Do not bulk-load the whole article catalog into output.