| name | sparq:validate |
| description | Validating existing E2E tests against current UI, codebase, and designs. Detecting broken selectors, stale flows, UI drift, and technical freshness issues. Use when: (1) tests exist but may be outdated, (2) after UI refactoring, (3) periodic health checks, (4) checking if tests still match current state. |
| audience | qa |
Validate Existing Tests
Config, version check, pattern rules, and E2E code generation preamble per claude/rules/skills.md. Additionally: grep {project.sourceRoot}/ for data-testid, labels, route definitions. Read validation-checklist.md and pattern-adherence.md before scanning.
Purpose
Validate existing E2E tests for UI drift, stale selectors, broken flows, and technical freshness. This skill focuses exclusively on whether tests match the current state of the UI and codebase -- it does not handle requirement changes (use /sparq:sync for that).
When NOT to Use
- If requirements changed (new acceptance criteria, updated specs, Jira ticket updated): use
/sparq:sync instead — it performs diff analysis against updated requirements.
/sparq:validate checks for UI/codebase drift (broken selectors, stale flows, API changes) — it validates tests against the current state of the application, not against changed requirements.
When to Use
- Tests exist but may be outdated after UI changes
- After a UI refactoring or component library update
- Periodic health checks on test suite freshness
- Checking if selectors, flows, and assertions still match current state
- Before a release to verify test technical validity
Workflow
- Scan test files: Read test files from specified path or defaults (E2E directories per
e2e.structure.* from config). Parse each to extract: selectors, URLs, expected text, flow sequences, assertions.
- Fetch current UI state in parallel from available sources (skip unavailable; codebase always works). For >10 test files: launch parallel validation checks per
parallel-execution.md Pattern 3 (up to 6 Task agents). For <=10 files: run checks sequentially.
- Figma: component names, layout, text content, element hierarchy
- Codebase: component files (using
project.componentFileExtensions), data-testid attrs, route definitions (using project.routeDiscoveryPattern)
- Live browser: actual rendered DOM, element visibility (Playwright CLI, when
e2e.framework is playwright). Not available for Cypress; skip silently.
- Compare and detect mismatches:
- Selector drift:
data-testid removed/renamed in source
- Flow changes: steps reference removed pages/routes
- UI text mismatches: assertion text differs from Figma/codebase
- Deprecated patterns: tests using old utilities/removed helpers
- Dead tests: tests for removed features
- New untested features: new components/routes without test coverage
- Performance regressions: tests with excessive waits or redundant steps
- Delegate findings to sparq-test-validator agent for classification (pass tech context:
project.componentFileExtensions, e2e, project.sourceRoot, project.routeDiscoveryPattern, preferences.locatorPriority)
- Apply deterministic fixes automatically (selector renames, import path updates) with checkpoint before applying
- Report non-deterministic issues with recommendations for manual resolution
Severity Levels
- Critical: test will fail (broken selector, removed route, deleted component)
- Warning: test may fail (renamed text, changed flow, fragile selector)
- Info: style/best practice (deprecated pattern, missing assertion, performance suggestion)
Output
.sparq/validation/validation-report.md
Report includes: VF-{n} IDs for each finding, severity classification, auto-fix proposals for Critical findings, recommendations for Warning/Info findings.
Fallback Behavior
When sources are unavailable, degrade per degradation-strategy.md. Codebase grep always works as baseline. If Figma unavailable: skip visual comparison, note gap. If Playwright not installed: skip live browser checks, rely on codebase analysis.
When findings are gathered in parallel, merge by severity (Critical first), assign unified VF-{n} IDs, deduplicate findings with identical selectors.
If multiple sources disagree on a selector or text value, flag as Warning with both values shown for user resolution.
<done_criteria>
- All spec files in the target path scanned
- Validation report generated at
.sparq/validation/validation-report.md with VF-{n} IDs
- Critical findings have auto-fix proposals
- Findings classified by severity (Critical/Warning/Info)
- Checkpoint presented before applying any fixes
- Unavailable sources documented as gaps in report
</done_criteria>
References
claude/skills/sparq-shared/references/validation-checklist.mdclaude/skills/sparq-shared/references/pattern-adherence.mdclaude/skills/sparq-shared/references/parallel-execution.mdclaude/skills/sparq-shared/references/degradation-strategy.mdclaude/skills/sparq-shared/references/config-schema.md
Examples
/sparq:validate e2e/
-> Scans 12 test files across all spec directories
-> Fetches current state (codebase + Figma + Playwright CLI in parallel)
-> CHECKPOINT: 3 Critical, 5 Warning, 2 Info findings
-> User picks "B" (critical only) -> applies 3 critical auto-fixes
-> Re-validates (cycle 1): 0 Critical, 5 Warning, 2 Info remaining
-> Output: .sparq/validation/validation-report.md
/sparq:validate e2e/specs/auth/
-> Scans 3 test files in auth directory only
-> Finds 1 Critical (broken selector), 2 Warning (changed text)
-> CHECKPOINT: presents findings with auto-fix proposals
-> User approves all -> applies fixes
-> Output: .sparq/validation/validation-report.md
After validation completes, consider running /sparq:sync if requirements may have also changed.
Complementary tool: Run sparq lint {e2e-directory}/ before or after validate for fast structural checks (locator quality, flaky patterns, assertion coverage) — deterministic, CI-compatible, zero model inference. /sparq:validate handles semantic/UI drift; sparq lint handles code-quality patterns.
