SLDS Quality Audit

Audit Lightning Web Components for SLDS compliance and produce an automated scorecard plus a required manual review gate. Combines SLDS linter output with supplementary static analysis to catch what the linter misses.

Scope

Also valid for: auditing SLDS compliance across a project or component set, and before/after quality comparison after making changes.

Not for:

• Fixing linter violations — use design-systems-slds2-migrate instead
• Building new components — use design-systems-slds-apply instead
• Just running the linter — run npx @salesforce-ux/slds-linter@latest lint . directly
• Full WCAG accessibility audit — this skill checks attribute presence only (labels, alt text, focus indicators), not contrast ratios, keyboard flows, or screen reader behavior
• Framework-specific template auditing beyond .css, .html, and .js files — JSX/TSX/Vue/Svelte outputs need additional manual review

Quality Validation Process

1. Run SLDS Linter     → Collect violation counts (linter's job)
2. Run Analyze Script  → Check what linter doesn't cover (supplementary)
3. Agent Review        → Required manual review gate
4. Score & Grade       → Compute automated score + final recommendation
5. Generate Report     → Produce formatted scorecard

Step 1: Run SLDS Linter

Run the linter to collect baseline violation data:

npx @salesforce-ux/slds-linter@latest lint <component-path> 2>&1

Count violations by rule. These feed directly into the Linter Compliance score:

Linter Compliance Score = 100 - (total_violations × 10), minimum 0.

If the linter is unavailable (no Node.js, no network access, CI sandbox restrictions): skip this step, note "Linter not run" in the report header, mark Linter Compliance as N/A, and compute the Overall score using the remaining 4 categories renormalized to 100%:

Overall (linter unavailable) = (Theming × 0.29) + (Accessibility × 0.29)
                              + (CodeQuality × 0.21) + (ComponentUsage × 0.21)

Step 2: Run Supplementary Analysis

Run the analyze script to catch issues the linter doesn't cover. The bundled analyzer scans .css, .html, and .js files only:

node scripts/analyze-quality.cjs <component-path>

The script outputs JSON with findings organized by severity. It checks:

CSS Checks (linter-complementary)

JS Checks

HTML Checks

Hook Pairing Validation

The script checks that background/foreground hooks are semantically paired:

surface-* backgrounds     → on-surface-* text
surface-container-* bg    → on-surface-* text
accent-* backgrounds      → on-accent-* text
accent-container-* bg     → on-accent-* text

Limitation: Hook pairing is checked at the file level, not per-selector. A file with surface-1 in .classA and on-accent-1 in .classB would pass because both surface and accent families are present. Review pairing correctness per-selector during manual review (Step 3).

Invented Hook Detection (T051)

The script cross-references every --slds-g-* token in CSS against hooks-index.json. Any hook not found in metadata is flagged as critical — this catches the most common agent mistake of inventing hooks from naming patterns.

Step 3: Agent Manual Review

These checks require understanding the component's purpose and cannot be automated reliably. Review each and classify findings as either:

• Blocking — incorrect blueprint structure, missing required states, or semantic/interaction issues that make the component not production-ready
• Advisory — worthwhile improvements that do not block shipping on their own

Manual review findings are not automated, but they do affect the final recommendation. Do not report an automated grade as the only verdict.

Step 4: Calculate Automated Scores and Final Recommendation

Component Complexity

Before scoring, classify the component to give the score context:

Include the complexity classification in the report header. This prevents misreading a "B" on a 1000-line component vs. a "B" on a 20-line component.

Automated Scoring Formula

Category Score = 100 - (critical_issues × 10) - (warnings × 3) - (info × 1)
Minimum score: 0

Categories and Weights

Automated Overall Score

Overall = (Linter × 0.30) + (Theming × 0.20) + (Accessibility × 0.20)
        + (CodeQuality × 0.15) + (ComponentUsage × 0.15)

Automated Grade Thresholds

Manual Review Gate

After computing the automated score, apply the manual review outcome:

Final Recommendation Rules

Use both the automated score and the manual review gate:

Step 5: Generate Quality Report

Use the template in [report-format.md](references/report-format.md) to produce the final report. Default to the compact format for initial output and expand sections on request.

The report includes:

• Executive summary with automated grade and final recommendation
• Manual review gate outcome (Pass, Advisory, or Blocking)
• Scores by category with visual indicators
• Detailed findings organized by severity
• Specific code locations and recommendations
• Checklist of required actions

Quick Validation Mode

For a rapid quality check without full analysis:

1. Run linter: npx @salesforce-ux/slds-linter@latest lint <path>
2. Count violations by type
3. Report summary only

Quick Quality Check: <component-name>
─────────────────────────────────────
Linter Violations:
  • Class Override:     0
  • Deprecated Tokens:  3
  • Hardcoded Values:   5

Quick Automated Grade: C (estimated)
Run full validation for detailed report.

Edge Cases and False Positives

If a check produces a false positive, note it in the report as "suppressed" with justification rather than silently dropping it.

References

• [Quality Checks](references/quality-checks.md) - Complete list of all quality checks with detection patterns
• [Report Format](references/report-format.md) - Quality report template and formatting guide
• [Analyze Script](scripts/analyze-quality.cjs) - Automated analysis for linter-complementary checks
• design-systems-slds2-migrate skill - How to fix linter violations
• design-systems-slds-apply skill - Guide for building new components with correct patterns