/audit - Code Review Orchestrator

Orchestrate specialized review agents with confidence-based filtering.

Execution

Task({
  subagent_type: "audit-orchestrator",
  description: "Comprehensive code review",
  prompt: `...`,
});

Review Agents (15 total)

Core (8)

Agent	Focus
structure-reviewer	DRY, coupling
readability-reviewer	Clarity, naming
type-safety-reviewer	TypeScript coverage
silent-failure-reviewer	Empty catch, Promise
design-pattern-reviewer	Pattern consistency
progressive-enhancer	CSS-first solutions
testability-reviewer	Test design, coverage
root-cause-reviewer	Root cause analysis

pr-review-toolkit (4)

Agent	Focus
silent-failure-hunter	Error handling
comment-analyzer	Documentation quality
type-design-analyzer	Type invariants
code-simplifier	Simplification

Production (3)

Agent	Focus
security-reviewer	OWASP
performance-reviewer	Bundle, render
accessibility-reviewer	WCAG, ARIA

Confidence Markers

• [✓] >0.8 - Verified with code evidence
• [→] 0.5-0.8 - Inferred with reasoning
• [?] <0.5 - Not included

Output Format

Review Summary
- Files: [count] | Critical [X] / High [X] / Medium [X]

## Critical Issues
[issues with file:line]

## Medium Priority
[issues with reasoning]

Recommended Actions
1. Immediate [✓]
2. Next Sprint [→]

IDR Update

After review, append /audit section to IDR with:

• Review summary
• Issues & actions
• Recommendations applied

Usage

/audit                    # Full review
/audit "src/components"   # Target scope

Next Steps

• Critical → /fix
• Refactoring → /think → /code