Code Review Skill
Orchestrates code review using the code-reviewer agent.
Purpose
Provides comprehensive code review with dual perspective:
- •Technical: Code quality, maintainability, SOLID principles
- •Functional: Ticket requirements, acceptance criteria, edge cases
Inputs
Required context:
- •Ticket content (ticket.md or inline)
- •Implementation plan (plan.md or inline)
- •Git diff of changes
Process
Step 1: Gather Context
- •Read ticket from
.claude/feature/{ticket-id}/ticket.md - •Read plan from
.claude/feature/{ticket-id}/plan.md - •Get git diff:
bash
git diff {base-branch}...HEAD
Step 2: Load Agent
Load ~/.claude/agents/code-reviewer.md and apply its instructions.
Step 3: Perform Review
Execute the agent's review process:
- •Functional completeness check
- •Code quality analysis
- •Maintainability assessment
- •Codebase consistency check
Step 4: Generate Report
Create review report in .claude/feature/{ticket-id}/review.md
Step 5: Present Results
Display summary:
- •Total issues found
- •Breakdown by severity
- •Recommended action (APPROVED/NEEDS_CHANGES/BLOCKED)
Configuration
From .claude/ticket-config.json:
json
{
"review": {
"enabled": true,
"auto_fix": false,
"severity_threshold": "important",
"model": null,
"block_on_critical": true
}
}
- •
enabled: Run review phase - •
auto_fix: Automatically apply suggested fixes - •
severity_threshold: Minimum severity to report (critical|important|minor) - •
model: Override model for different analysis perspective - •
block_on_critical: Prevent PR creation if critical issues exist
Status Values
- •
pending: Review not yet started - •
in_progress: Review in progress - •
completed: Review finished successfully - •
skipped: Review skipped (disabled or no implementation) - •
failed: Review encountered an error
Integration
Invoked by:
- •
/resolveworkflow (STEP 10) - •
/review-codestandalone command