polish-sweep

/polish-sweep

$ARGUMENTS

---

This is the "last 15%" - issues that users feel but can't articulate. Missing feedback, inconsistent behaviors, dead-end errors, and code patterns that lead to UX problems.

Focus on HIGH and MEDIUM severity findings. Low severity is polish-on-polish - nice to have but not blocking.

• No argument or "all": Scan components/ and app/ directories
• "components": Scan only components/ directory
• "app": Scan only app/ directory
• Specific path: Scan the provided directory

Note: Default directories are optimized for Next.js/React projects. For other frameworks (Vue, Angular, etc.), specify a custom path like src/components.

Communicate the scope to each agent.

Spawn all 4 reviewer agents simultaneously using the Task tool. Each receives the scope and returns structured findings with file paths, line numbers, severity, and suggested fixes.

Agents:

• ux-clarity-reviewer — Missing feedback (loading, success, error, empty states)
• ux-consistency-reviewer — Behavioral inconsistency (tooltips, animations, patterns)
• recovery-reviewer — Dead-end error paths (missing retry, data loss, no escape)
• code-consistency-reviewer — Code pattern variance (prop naming, conventions)

Aggregation

Combine agent outputs: deduplicate same-file+line findings, sort by severity (HIGH → MEDIUM → LOW), group by file, track pattern frequency.

Report Format

Format findings for actionability:

# Polish Sweep Report

**Scope:** components/, app/ **Found:** X issues (Y high, Z medium)

## High Priority

### path/to/file.tsx

- **[clarity]** Line 45: Async operation without loading state → Add isLoading state,
  show spinner during fetch

- **[recovery]** Line 78: Error caught but no user feedback → Show error toast with
  retry option

## Medium Priority

### path/to/other-file.tsx

...

## Pattern Summary

| Issue Type            | Count | Files |
| --------------------- | ----- | ----- |
| Missing loading state | 5     | 4     |
| Inconsistent tooltip  | 3     | 3     |
| Dead-end error        | 2     | 2     |