/do-review

Do Review: Code Review via comprehensive-review

Trigger a code review on your recent changes using the comprehensive-review plugin. This is Step 3 in the Development Flow — review before committing.

Usage

/do-review                    # Review all uncommitted changes
/do-review HEAD~3..HEAD       # Review a 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 (alias for --security)

Tasks

1. Validate arguments and determine review scope

Parse arguments into **scope** and **agent flags**:

• **Scope args** (mutually exclusive): commit range (e.g. `HEAD~3..HEAD`), `--branch`, or no args
• **Agent flags** (optional, combinable with any scope): `--security`, `--full`

**Unrecognized argument** (not a scope arg or agent flag):

Usage: /do-review [HEAD~N..HEAD | --branch] [--security | --full]

Stop here. Do not proceed.

Based on scope:

| Invocation | Scope | Git command | |-----------|-------|-------------| | `/do-review` (no scope args) | All uncommitted changes | `git diff HEAD` | | `/do-review HEAD~3..HEAD` | Specific commit range | `git diff HEAD~3..HEAD` | | `/do-review --branch` | Current branch vs main | `git diff main...HEAD` |

Run the appropriate git diff command.

**If no changes found:**

Nothing to review. No changes detected.

Stop here. Do not proceed.

**If large changeset (20+ files or 500+ lines):**

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

If user narrows scope, re-run with the adjusted scope.

2. Prepare review context

Gather context for the review agents:

1. **Get the diff** — output of the git diff command from step 1 2. **List changed files** — extract file paths from the diff 3. **Load coding standards** — for each changed file, check if a matching coding standards skill exists:

• Identify file extensions in the changeset
• Read matching skills from `~/.claude/skills/` (follow the Skill Loading table in global CLAUDE.md)
• Collect relevant standards content

4. **Read project context** — tech stack and current task from project CLAUDE.md (About section, Tech Stack, Current Status)

3. Run review agents

Determine which agents to spawn based on flags:

| Flag | Agents | |------|--------| | (none) | architect-review + code-reviewer | | `--security` | architect-review + code-reviewer + security-auditor | | `--full` | architect-review + code-reviewer + security-auditor |

Spawn all agents **in parallel** via the Task tool. Don't wait for one to finish before spawning the next.

**Agent 1: comprehensive-review:architect-review**

Review the following code changes from an architectural perspective.

## Project Context
Tech Stack: [from project CLAUDE.md]
Current task: [from Current Status table, if available]

## Changes
[git diff output]

## Changed Files
[list of file paths — read these for full context around the changes]

## Coding Standards
[relevant standards from loaded skills]

Focus on: system design, architecture patterns, SOLID principles, scalability, dependency management.
Be actionable — suggest specific improvements, not general observations.
Skip praise; focus on what can be improved.
If everything looks good, say so briefly.

**Agent 2: comprehensive-review:code-reviewer**

Review the following code changes for code quality and maintainability.

## Project Context
Tech Stack: [from project CLAUDE.md]
Current task: [from Current Status table, if available]

## Changes
[git diff output]

## Changed Files
[list of file paths — read these for full context around the changes]

## Coding Standards
[relevant standards from loaded skills]

Focus on: code quality, maintainability, error handling, best practices, production reliability, test coverage gaps.
Be actionable — suggest specific improvements, not general observations.
Skip praise; focus on what can be improved.
If everything looks good, say so briefly.

**Agent 3 (only with --security or --full): comprehensive-review:security-auditor**

Perform a security audit on the following code changes.

## Project Context
Tech Stack: [from project CLAUDE.md]
Current task: [from Current Status table, if available]

## Changes
[git diff output]

## Changed Files
[list of file paths — read these for full context around the changes]

Focus on: security vulnerabilities, OWASP top 10, input validation, authentication/authorization, secrets handling, injection attacks.
Be actionable — suggest specific fixes, not general warnings.
Skip praise; focus on what can be improved.
If everything looks good, say so briefly.

**Spawn configuration:**

• Agent 1: `subagent_type`: `comprehensive-review:architect-review`
• Agent 2: `subagent_type`: `comprehensive-review:code-reviewer`
• Agent 3: `subagent_type`: `comprehensive-review:security-auditor`
• Let agents read the full files themselves — don't paste entire file contents

**If a review agent fails because the plugin is not installed:**

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

Install it: /claude-code-setup → External Plugins
Or manually: claude plugin marketplace add wshobson/agents && claude plugin install comprehensive-review@claude-code-workflows

Stop here.

**If a review agent fails for other reasons:**

• If at least one agent completed: present the available results, note which agent(s) failed
• If all agents failed: show the error and stop

Review could not be completed: [error]
You can retry with /do-review or request a manual review.

4. Present findings

Collect results from all agents. Present consolidated findings grouped by perspective:

## Architecture Review (architect-review)
[findings]

## Code Quality Review (code-reviewer)
[findings]

## Security Audit (security-auditor) ← on