Reviewer Common Patterns

Common patterns referenced by all reviewer agents.

1. Confidence Marker System

Marker	Confidence	Usage
[✓]	>0.8	Verified (file:line required)
[→]	0.5-0.8	Inferred (state reasoning)
[?]	<0.5	Hypothesis (needs confirmation)

Evidence requirement: All findings must include file:line, code snippet, and impact.

Source: @../../rules/core/AI_OPERATION_PRINCIPLES.md

2. Standard Review Process

1. Context Discovery: Identify project conventions, existing patterns
2. Assessment: Apply skill knowledge, detect patterns, evaluate confidence
3. Filtering: Exclude below threshold (0.7), document evidence

3. Integration Patterns

When findings overlap between agents:

• Primary owner: Agent whose domain is central
• Secondary: Cross-reference related findings
• No duplicates: Same issue not reported by multiple agents

4. Browser/MCP Verification (Optional)

Applies to: accessibility-reviewer, performance-reviewer

Use When	Skip When
Complex interactions	Static HTML/CSS
Custom ARIA	No dev server
Performance profiling	Code analysis sufficient

Fallback: If MCP unavailable, code-only analysis with [→] marker.

5. JP/EN Translation Handling

Files under .ja/ are Japanese translations.

Review	Skip
Structure match	Content comparison
YAML match	Translation differences
Link validity	Localized formats

Source: @../../rules/guidelines/JP_EN_TRANSLATION_RULES.md