form-ux-best-practices（形成用户体验最佳实践）

Form UX Best Practices

Operating Mode

• Run in text-only mode by default.
• Avoid image-led critiques, dashboards, or visual redesign directions unless explicitly requested.
• Prioritize decisions that improve completion quality, error recovery, accessibility, and trust.

Inputs Required

Collect these inputs before reviewing:

• Form context: signup, checkout, settings, lead-gen, or other.
• Primary business goal and user goal.
• Current form source: spec, HTML, React, screenshots transcribed to text, or mixed.
• Field list with required/optional status.
• Current validation behavior (inline, blur, submit, server-side).
• Target devices (desktop, mobile, both).

Intake Questions

Ask only if essential inputs are missing. If essentials are present, proceed without questions.

• What is the exact decision this form should support (create account, complete purchase, update settings, capture lead)?
• Which fields are mandatory to complete that decision?
• What happens after submit (success path, retry path, abandonment path)?
• Are there legal or policy constraints (age checks, billing details, consent, regional rules)?

Deterministic Review Workflow

1. Define task-critical outcome: state one sentence for user success and one sentence for business success.
2. Map friction points: identify every extra field, ambiguous label, avoidable decision, and interruption.
3. Audit field architecture: check sequence, grouping, required/optional clarity, and progressive disclosure.
4. Audit validation behavior: check timing, specificity, recovery path, and consistency between inline and submit states.
5. Audit accessibility semantics: labels, instructions, focus order, keyboard use, and error associations.
6. Audit mobile ergonomics: input type, inputmode, keyboard hints, spacing, touch target comfort, and scroll/focus behavior.
7. Audit trust signals: explain why sensitive fields are requested and place privacy assurances at decision moments.
8. Produce outputs using assets/report_template.md exactly.

Reference Loading

• Load references/canon.md only when you need principle-level rationale or a tie-breaker between competing form patterns.
• Prefer the deterministic workflow above for execution; use canon references to justify recommendations, not to replace direct analysis.

Accessibility Checks

• Ensure every control has a programmatic name from a visible label, not placeholder-only text.
• Ensure label and control association via for/id or explicit wrapper label pattern.
• Ensure helper text and errors are referenced by aria-describedby where applicable.
• Ensure invalid fields set aria-invalid="true" after validation fails.
• Ensure error summary (if used) links to each invalid field and receives focus after failed submit.
• Ensure tab/focus order matches visual/logical order.
• Ensure keyboard-only completion is possible without pointer interaction.

Mobile Checks

• Use the correct keyboard hints (type, inputmode, autocomplete, autocapitalize).
• Keep label visibility persistent while typing.
• Avoid multi-column field arrangements that compress touch targets on narrow screens.
• Keep helper/error text close to field and visible without horizontal scrolling.
• Avoid forced context-switch loops (field blur -> jump -> keyboard close/open thrash).

Validation Strategy

• Validate required presence on submit, not on first keystroke.
• Validate format either on blur or when enough characters exist to infer intent.
• Reserve real-time validation for high-value, low-noise cases (for example: password strength meter, username availability).
• Show one clear recovery action per error.
• Keep inline error copy consistent with submit-time error summary copy.
• Use error summary for multi-error forms or long forms where first error may be off-screen.

Field Design Rules

• Prefer explicit labels above fields; do not rely on placeholders as labels.
• Mark optional fields explicitly; keep required as default unless product policy requires explicit required badges.
• Request only minimum data needed for the current decision.
• Group related fields and follow user mental model order (for checkout: contact -> delivery -> payment).
• State constraints before input when predictable (format, length, character rules).

Microcopy Rules

• Use direct, specific phrasing that names the issue and next action.
• Avoid blame language and vague statements like "Invalid input".
• Mention accepted format when format errors are common.
• Keep tone neutral and recovery-oriented.
• Keep labels and error terms consistent (same field name in label, helper, and error).

Privacy and Trust Cues

• Explain why sensitive information is needed at the moment it is requested.
• Place trust cues near commitment actions (submit/place order), not buried in footer text.
• Clarify consent scope and frequency for communications fields.
• Separate mandatory operational consent from optional marketing consent.

Conversion and Friction Rules

• Remove non-essential fields from initial flow.
• Use progressive disclosure for secondary details.
• Prefer smart defaults when safe and reversible.
• Preserve entered values after failed submit.
• Minimize redundant confirmation fields unless risk justifies it.

Prioritization Rubric (P0/P1/P2)

• P0: Blocks completion, creates legal/compliance risk, or makes the form unusable for assistive tech.
• P1: Causes significant confusion, avoidable errors, or measurable abandonment risk.
• P2: Improves clarity, speed, and polish but does not block completion.

Output Format

Always fill assets/report_template.md sections in this exact order:

1. Top Issues (P0/P1/P2)
2. Field-by-field rewrite
3. Validation & error messaging spec
4. Accessibility checklist
5. Ship-ready checklist
6. Optional: Ready-to-ship snippet

Optional Ship Snippet Mode

When asked to produce implementation-ready guidance:

• Output either semantic HTML or React pseudo-code.
• Include labels, helper text hooks, error bindings, and autocomplete attributes.
• Include submit-time summary pattern when there are 2+ possible simultaneous errors.
• Keep snippet focused on form behavior and semantics, not visual styling.

Optional Static Audit Script

Use scripts/form_audit.py for quick static checks on HTML forms.

python3 scripts/form_audit.py /path/to/form.html

Script checks:

• Missing label mapping
• Duplicate IDs
• Missing name attributes
• Placeholder used as label
• Missing autocomplete for common fields
• Likely input type mismatches