comprehensive-review

AI-powered code review with architecture, quality, and security agents.

Overview

Field	Value
Marketplace	claude-code-workflows
Source	wshobson/agents
License	MIT
Default	Pre-selected (recommended)

Why Recommended?

This plugin integrates with our development workflow. The /do-review command uses its agents for multi-perspective code review.

1. SPECIFY
         │
         ▼
2. IMPLEMENT
         │
         ▼
3. CODE REVIEW  ◄──── comprehensive-review
   - Run /do-review
   - Incorporate feedback

Agents

The plugin provides 3 specialized review agents:

Agent	Focus	Used by
`architect-review`	Architecture, design patterns, SOLID, scalability	`/do-review` (always), `/design --review`
`code-reviewer`	Code quality, maintainability, error handling, best practices	`/do-review` (always)
`security-auditor`	OWASP, vulnerabilities, auth, injection, secrets	`/do-review --security` or `--full`

Agents run in parallel for fast reviews.

Installation

Via Installer

./install.sh        # Fresh install
./install.sh --add  # Add to existing

comprehensive-review is pre-selected by default.

Manual

# Add marketplace
claude plugin marketplace add wshobson/agents
 
# Install plugin
claude plugin install comprehensive-review@claude-code-workflows

Usage

Via /do-review (recommended)

The /do-review command is the primary way to trigger this plugin:

/do-review                    # 2 agents: architect-review + code-reviewer
/do-review --security         # 3 agents: + security-auditor
/do-review --full             # 3 agents (alias for --security)
/do-review --branch           # Review branch vs main
/do-review --branch --full    # Full review before PR

It automatically gathers the diff, loads coding standards, and spawns agents in parallel.

Via /design —review

The /design --review command uses architect-review for design feedback.

Direct Request

"Review this PR for architectural issues"
"Analyze the security of this module"
"Check this code for best practices"

Via Task Tool

The plugin provides subagents that can be spawned via the Task tool:

subagent_type: "comprehensive-review:architect-review"
subagent_type: "comprehensive-review:code-reviewer"
subagent_type: "comprehensive-review:security-auditor"

Review Output

Each agent provides focused feedback:

architect-review — Design concerns, dependency issues, pattern violations
code-reviewer — Quality improvements, maintainability, error handling gaps
security-auditor — Vulnerabilities, OWASP findings, auth/validation issues

Results are grouped by perspective for easy reading.

Integration with Workflow

After implementing a feature:

Run /do-review (or /do-review --full for sensitive changes)
Agents review the changes in parallel
Feedback is incorporated
Code is committed

This catches issues before they get committed.

See /do-review for details.

Overview document-skills