gstack/test/fixtures/plan-count-design-questionless-report.md

Settings Page UI Redesign — Design Review

Context

Reviewing the Settings Page UI Redesign plan against the checked-in DESIGN.md. The plan documents existing behavior in detail and calls out 5 implementation gaps explicitly. This review evaluates all 7 design dimensions, confirms each flagged gap, identifies 2 additional gap areas, and proposes concrete fixes traceable to DESIGN.md tokens.

DESIGN.md exists and is comprehensive. All ratings calibrate against it.

Initial rating: 5/10 — Behavior spec is thorough. Design token gaps leave the plan incomplete as an implementation guide. A 10/10 plan has every visual decision traceable to a specific token or measurement with no ambiguity for the implementer.

---

Dimension 1: Information Architecture — 8/10

Current state. The plan specifies h1/h2 heading hierarchy, main landmark, fieldsets with aria-labelledby, and a fixed DOM order. Sections are fixed: Profile and Notifications. Default values for new accounts are defined. Read failure state (Retry) is covered.

What a 10 looks like. The h1 description text is specified verbatim. Every string visible in the UI is either specified or explicitly delegated to an existing source.

Gap. The "short description" beneath the h1 is not specified. The plan says the page header "contains the title, a short description" but gives no copy. For an existing form, this string already exists — the plan should confirm it's unchanged or specify the exact text.

Fix. Add one line: "The description text is unchanged from the current form" or provide the verbatim string. No structural change needed.

After fix: 9/10

---

Dimension 2: Visual Hierarchy — 5/10

Current state. Four buttons in the header action group share the same size, weight, and color. No visual signal tells the user which is primary. InlineStatus position (between actions and form) is correct.

What a 10 looks like. The user's eye lands on Save first without hesitation. The three secondary actions read as secondary. Section headings outweigh field labels. Every element's visual weight matches its semantic importance.

Gap (flagged in plan). Save button must be visually primary. DESIGN.md specifies: Save = #1d4ed8 fill with white text. Reset, Cancel, Export = neutral ghost buttons. The confirmation dialog explains destructive intent for Reset; visual weight reduction there is safe.

Additional gap. Ghost button appearance is specified by name only — "neutral ghost buttons" — with no color tokens for border, text, or hover state. An implementer must guess. This is a common source of inconsistency between developers.

Fixes.

1. Apply primary style to Save: background-color #1d4ed8, color #ffffff. This is the only filled button on the page.
2. Specify ghost style for Reset/Cancel/Export: border 1px solid #d1d5db, color #374151, hover background rgba(0,0,0,0.04). All pairs exceed 4.5:1 on white.
3. Add a note: "Export is a ghost button, not a link, because it performs an action (download) rather than navigating."

After fixes: 9/10

---

Dimension 3: Typography — 5/10

Current state. Three font sizes appear across the form — 14px, 16px, 18px — creating a weak and inconsistent hierarchy. The h1 page title likely uses a fourth size not specified in the gap description.

What a 10 looks like. Two sizes. One for all body copy, labels, and helper text. One for section headings. Font weights explicitly specified for each text role so no implementer must guess between 400, 500, and 600.

Gap (flagged in plan). DESIGN.md specifies 16px for body/labels/helper text and 20px for section headings (h2). The implementation introduces 14px and 18px, both outside the approved scale.

Additional gap. Font weights are not specified in the plan or DESIGN.md. The implementer faces: Should h2 headings be 600 or 700? Should labels be 500 or 400? What weight for button text, helper text, status text? These choices produce visibly different results and should not be left to implementation judgment.

Fixes.

1. Replace all 14px with 16px (likely helper text or secondary labels).
2. Replace all 18px with 20px if on section headings, or 16px if on body-role text.
3. Add font weight specification to the plan:
   • h1 (page title): font-weight 700
   • h2 (section headings): font-weight 600
   • Form labels: font-weight 500
   • Helper text, status text, descriptions: font-weight 400
   • Button text: font-weight 500

After fixes: 9/10

---

Dimension 4: Color / Contrast — 5/10

Current state. The error message uses unspecified red text on a light pink background at approximately 3:1 contrast. WCAG AA requires 4.5:1 for normal text. This fails.

What a 10 looks like. Every color value is a specified hex token. Every text/background pairing has a verified contrast ratio in the plan. No state communicates meaning through color alone (icon and text accompany color).

Gap (flagged in plan). Error colors fail WCAG AA. DESIGN.md specifies #991b1b on #fef2f2 with an icon. Calculated contrast for #991b1b on #fef2f2: approximately 8.9:1 (well above AA 4.5:1 and AAA 7:1). The DESIGN.md-specified pair is correct.

Additional gaps.

1. Ghost button text color (#374151) and border color (#d1d5db) need explicit specification (also noted in Dimension 2). Both must be verified against white background: #374151 on white is approximately 8.0:1; #d1d5db is decorative border, not text, so WCAG AA does not apply but Non-text contrast (3:1) does — this should be checked.
2. Disabled Save button color during pending state is not specified. A disabled style must not rely solely on low opacity (fails Non-text contrast at thin opacities) and must have an explicit aria-disabled or disabled attribute.
3. Error icon color is not specified — it should match #991b1b to avoid a two-color error signal that fragments the semantic unit.

Fixes.

1. Error text: #991b1b. Error surface: #fef2f2. Error icon: #991b1b. All three explicitly stated in the plan's implementation tasks.
2. Ghost button: border #d1d5db (contrast against white: ~1.6:1 — confirm this is a decorative border and not the only focus indicator; the 2px #1d4ed8 focus ring covers keyboard users). Text: #374151.
3. Disabled Save during pending: use aria-disabled=true on the button element. Visual appearance: background #93c5fd, text #1e40af (or the existing component's disabled token if one exists). Do not rely on opacity alone.

After fixes: 9/10

---

Dimension 5: Spacing / Rhythm — 6/10

Current state. Three spacing values appear without consistent assignment rules: 24px in some places, 32px in others, 16px in a third. Some placements are correct per DESIGN.md (24px field groups, 32px sections), but the appearance of a third value suggests drift from the 8px base scale.

What a 10 looks like. Every spacing value is a multiple of 8px. The implementer reads the spec and knows exactly which spacing applies where without inference. No two sections of the form use different gap values for the same semantic relationship.

Gap (flagged in plan). DESIGN.md specifies the 8px base scale: sections = 32px, field groups = 24px, label-to-input = 8px. The rogue 16px value likely appears where 8px (label-to-input) is specified, or in an un-spec'd relationship (e.g., action row to InlineStatus) where the implementer defaulted to their judgment.

Additional gap. Several spacing relationships are unspecified in the plan and DESIGN.md:

• Gap between h1 block (title + description) and action row
• Gap between action row and InlineStatus
• Gap between InlineStatus and first fieldset
• Vertical padding inside fieldsets (around the group of fields)
• Gap between a field label and its helper text
• Gap between a field and its inline error message

Fixes.

1. Audit existing implementation for any gap value not on the 8px scale; replace with the nearest 8px multiple.
2. Add the following to the plan's spacing spec (all on the 8px scale):
   • h1 block → action row: 24px
   • Action row → InlineStatus: 16px
   • InlineStatus → first fieldset: 24px
   • Inside fieldset, vertical padding top/bottom: 24px
   • Field label → helper text: 4px
   • Field → inline error message: 4px
3. Label-to-input remains 8px per DESIGN.md.

After fixes: 8/10 — A 10/10 would enumerate dialog internal spacing too; leaving that to the existing ConfirmationDialog component's own spec is an acceptable delegation.

---

Dimension 6: Motion / Feedback States — 5/10

Current state. The Save action takes 2-5 seconds with no visual feedback. The page appears frozen during the operation. Users cannot distinguish between a slow save and a crashed interface.

What a 10 looks like. Every async operation has a defined pending state. Every state transition is perceptible to all users (motion and non-motion). The minimum spinner display time prevents flash. Reduced-motion users receive equivalent status information.

Gap (flagged in plan). DESIGN.md specifies: inline spinner beside "Saving…" text inside the disabled Save button, aria-busy=true, with reduced-motion support.

Additional gaps.

1. Minimum spinner display time is not specified. For saves completing in under 300ms, the spinner flashes briefly and vanishes — perceptually worse than no spinner. A floor of 500ms prevents the flash while still feeling responsive.
2. Dialog open/close transition is not specified. The plan references the existing ConfirmationDialog component. If that component has no animation, state it explicitly so the implementer does not add one ad hoc.
3. Error state appearance transition is not specified. Does the error summary animate in or appear immediately? Immediate appearance is correct (no motion delay for errors).

Fixes.

1. Save pending state:
   • Disable Save button, set aria-busy=true on the button element.
   • Show inline spinner (16px, #ffffff, rotating) beside "Saving…" text.
   • Reduced-motion: omit spinner, show "Saving…" text with a static icon or no icon.
   • Enforce minimum display time of 500ms: even if the server responds in 100ms, hold the pending state visible for 500ms before transitioning to success.
2. Dialog transitions: use existing ConfirmationDialog component defaults. No new animation. Explicitly note this in the implementation tasks.
3. Error appearance: immediate, no transition. Explicitly note this.
4. Export feedback: see Dimension 7.

After fixes: 8/10

---

Dimension 7: Interaction States & Accessibility — 7/10

Current state. The plan covers a wide range of states: initial load skeleton, new account defaults, read failure (Retry), editing (dirty state via "Unsaved changes"), pending save (gap), save success ("Saved at HH:mm"), save failure (retain edits + error), reset flow (confirmation), cancel flow (confirmation), and export. Accessibility is strong: heading hierarchy, ARIA roles, focus management, 44px targets, visible focus rings, dialog focus trap, and Escape key support.

What a 10 looks like. Every user action has a specified system response. No state is left for the implementer to decide. Validation timing is explicit. All async operations have defined pending states. Edge cases (47-char names, zero-length inputs, duplicate email) are either handled or explicitly deferred.

Gap 1: Export pending state is missing. The plan specifies Export failure and retry but not what happens while the JSON download is being prepared. Is the button disabled? Is there aria-busy? Is there a visual indicator? For a synchronous JSON export (common for client-side serialization), this is instantaneous. If that's the case, state it explicitly — the implementer should not add unnecessary loading UI for a <10ms operation.

Gap 2: Validation timing is not specified. The plan mentions field validation and aria-describedby for errors, and that Save is atomic. But when does validation fire? On blur? On submit? On keystroke? Each produces a different user experience. On-submit-only is consistent with the atomic save pattern and avoids premature errors before the user finishes typing.

Gap 3: Field character limits are not specified. Display name and Email fields have no specified maximum length. If the server enforces a limit (likely), the form should either show a character counter near the limit or surface a clear, specific validation error when exceeded. "Too long" is not specific; "Display name must be 100 characters or fewer (currently 143)" is.

Fixes.

1. Export pending: "Export is a synchronous client-side JSON serialization. The button disables for the duration of the download prompt (browser-native, < 50ms). No spinner or aria-busy needed. On failure (serialization error), show error in the inline status area per the existing error/retry pattern."
2. Validation timing: "Validate on submit only. After a failed submit, field-level errors appear beside each invalid field and in the linked error summary. Editing a field after an error clears that field's error immediately (on input event). Re-validation runs on the next submit attempt."
3. Field limits: "Display name: max 100 characters. Email: standard email format validation, max 254 characters (RFC 5321). Both validated server-side; client validation is progressive enhancement only."

After fixes: 9/10

---

Summary

Dimension	Before	After	Critical Fix
1. Information Architecture	8/10	9/10	Confirm h1 description copy
2. Visual Hierarchy	5/10	9/10	Apply primary style to Save; specify ghost tokens
3. Typography	5/10	9/10	Remove 14px/18px; specify font weights
4. Color / Contrast	5/10	9/10	Apply DESIGN.md error colors; specify disabled state
5. Spacing / Rhythm	6/10	8/10	Apply 8px scale; enumerate unspecified gaps
6. Motion	5/10	8/10	Save spinner with 500ms minimum; specify others
7. Interaction States	7/10	9/10	Export pending; validation timing; field limits

Blocking before implementation: Dimensions 2, 3, 4 (design token gaps will produce inconsistent implementations if unresolved).

Resolve before coding starts: Dimensions 5, 6, 7 (behavioral ambiguities will surface as bugs or review findings if unresolved).

Minor / low-risk: Dimension 1 (copy confirmation).

---

GSTACK REVIEW REPORT

Run	Status	Findings
plan-design-review	complete	7 dimensions reviewed, 13 findings across all dimensions
Outside voices	skipped	text-only mode per user request
Mockups	skipped	text-only mode per user request

VERDICT: NEEDS_REVISION — 5 of 7 dimensions rated below 7/10. Three dimensions (Visual Hierarchy, Typography, Color) have gaps that will produce inconsistent implementations. All fixes trace directly to DESIGN.md tokens; no design decisions require new choices. Apply the DESIGN.md-specified tokens and add the specificity items to the implementation plan.

UNRESOLVED DECISIONS:

• D1: Confirm h1 description copy is unchanged from current form, or specify verbatim string
• D2: Specify ghost button border/text/hover tokens (border #d1d5db, text #374151 recommended)
• D2: Confirm disabled Save visual tokens during pending state
• D3: Add font weight spec to plan (h1:700, h2:600, label:500, helper:400, button:500 recommended)
• D5: Enumerate unspecified spacing relationships (h1→actions, actions→status, status→fieldset)
• D6: Confirm 500ms minimum spinner display time; confirm no dialog transition animation
• D7: Confirm Export is synchronous (no pending state needed); specify validation fires on submit only; specify field character limits