Commands/do-review

/do-review

Code review on your recent changes using the comprehensive-review plugin. Step 3 in the Development Flow.

Usage

/do-review                    # Review all uncommitted changes
/do-review HEAD~3..HEAD       # Review specific commit range
/do-review --branch           # Review current branch vs main
/do-review --security         # Include security audit
/do-review --full             # Run all 3 review agents

What It Does

  1. Checks for the plugin — requires comprehensive-review
  2. Gathers changes — uncommitted, commit range, or branch diff
  3. Loads coding standards — matches file extensions to installed skills
  4. Spawns review agents — 2 by default, 3 with --security/--full
  5. Shows findings — grouped by perspective with actionable suggestions
  6. Offers to apply — incorporate feedback directly or pick specific items

Review Agents

AgentFocusWhen
architect-reviewArchitecture, design, SOLID, scalabilityAlways
code-reviewerCode quality, maintainability, best practicesAlways
security-auditorOWASP, vulnerabilities, auth, input validation--security or --full

Agents run in parallel — the review takes about as long as the slowest agent, not the sum.

Review Scopes

InvocationWhat gets reviewedWhen to use
/do-reviewStaged + unstaged changesBefore committing
/do-review HEAD~3..HEADSpecific commit rangeReviewing recent commits
/do-review --branchCurrent branch vs mainBefore creating a PR

Scope and agent flags combine freely: /do-review --branch --full

Example Session

You: /do-review

Claude: Reviewing uncommitted changes (4 files, ~120 lines)...

## Architecture Review (architect-review)

1. Dependency direction: utils/auth.py imports from handlers/
   — this inverts the dependency. Utils should not depend on handlers.

## Code Quality Review (code-reviewer)

1. process_items() has a nested loop that could be O(n*m).
   Consider using a dict for lookups.

2. Function getData doesn't follow snake_case convention
   per project standards.

Review complete. 2 agents, 3 findings total.

Incorporate feedback? [Yes / No / Pick specific items]

You: Yes

Claude: Applied all 3 suggestions.
        Run /do-review again to verify, or commit.

How It Fits the Workflow

1. SPECIFY (/design or CLAUDE.md)


2. IMPLEMENT (write code + tests)


3. REVIEW (/do-review)  ◄── you are here


   Back to 1. or /wrapup → /clear

/do-review fills the gap between implementing and committing. It automates what the Development Flow describes as Step 3.

What Gets Passed to Review Agents

The command automatically gathers:

  • The diff — what changed
  • Changed file list — agents read full files for context
  • Coding standards — from skills matching the file extensions
  • Project context — tech stack and current task from CLAUDE.md

You don’t need to provide any of this manually.

Large Changesets

For diffs over 20 files or 500+ lines, Claude will warn you:

Large changeset: 32 files, ~800 lines changed.
Review may be less thorough. Continue as-is, or narrow the scope?
[Continue] / [Specify files or range]

You can narrow with a commit range or review in smaller batches.

vs /design —review

/do-review/design —review
ReviewsCode changes (diffs)Design Records
WhenAfter implementingDuring design
FocusArchitecture, code quality, securityProblem clarity, story quality, completeness
Agentsarchitect-review + code-reviewer (+ security-auditor)architect-review only
ScopeGit changesRecord content

Prerequisites

Requires the comprehensive-review plugin. If not installed:

comprehensive-review plugin is not installed. This command requires it.

Install it: /claude-code-setup → External Plugins

Tips

  • Run after implementing but before committing — catches issues early
  • Use --branch before creating PRs for a comprehensive review
  • Add --full for security-sensitive changes (auth, API endpoints, data handling)
  • Coding standards are included automatically — no need to mention them
  • You choose what feedback to act on — nothing is applied without your say
  • Works well with /with-advisor: advisor during implementation, review after