How to Integrate the Code-Reviewer Agent into a Claude Code Development Workflow

The code-reviewer agent in the everything-claude-code (ECC) ecosystem automatically inspects every file change for security flaws and quality defects, blocking commits that contain CRITICAL or HIGH severity issues until they are resolved.

The everything-claude-code repository by affaan-m provides a framework for automating software development tasks within Claude Code. Learning how to integrate the code-reviewer agent into a development workflow with Claude Code ensures that every modification is validated against security and quality standards before it reaches the codebase.

Architecture of the Code-Reviewer Agent

Agent Definition and Tools

The behavior of the code-reviewer is defined in agents/code-reviewer.md. This file specifies the four core tools the agent employs: Read, Grep, Glob, and Bash. These tools enable the agent to inspect file contents, search for patterns across the codebase, and execute shell commands to gather context for its analysis.

Automatic Triggering via Hooks

ECC implements an agent orchestration system described in AGENTS.md. The hook system automatically invokes the code-reviewer agent immediately following any file write or edit operation. This automatic triggering ensures that developers receive instant feedback without needing to remember to initiate a review manually.

How the Review Process Works

Diff Analysis and Checklist Evaluation

When activated, the code-reviewer executes git diff --staged (or git diff if nothing is staged) to collect the full set of changes. The agent parses this diff and applies a comprehensive checklist that includes:

• Scanning for hardcoded secrets and API keys
• Detecting SQL injection patterns
• Identifying overly large functions
• Flagging missing error handling

The specific logic for these checks is documented in the review checklist section of agents/code-reviewer.md.

Severity-Based Blocking Logic

The commands/code-review.md file implements the decision logic that governs whether a commit may proceed. If the agent detects any CRITICAL or HIGH severity issues, the commit is automatically blocked. The developer must resolve these issues and re-run the review before the Claude Code toolchain allows the git commit operation to proceed.

Integrating the Code-Reviewer into Your Workflow

Manual Invocation with Slash Commands

Developers can trigger the code-reviewer on demand using the /code-review slash command within Claude Code. This command executes the review logic defined in commands/code-review.md and returns a formatted report grouped by severity, including file paths, line numbers, and suggested fixes.

Example output:

[CRITICAL] Hardcoded API key in source
File: src/api/client.ts:42
Issue: API key "sk-abc…" exposed in code.
Fix: Move to environment variable and update .gitignore

Automatic Review After File Edits

The most seamless integration relies on ECC's hook system. When you edit a file such as src/utils/formatDate.ts, the system automatically:

1. Detects the file change
2. Invokes the code-reviewer agent
3. Returns feedback directly in the Claude Code chat window

If the agent reports a HIGH severity issue, such as a function exceeding 50 lines, you can refactor the code immediately. Upon saving, the agent re-evaluates the changes and clears the block once the issue is resolved.

Orchestrated Multi-Agent Pipelines

For complex features, integrate the code-reviewer into a multi-agent pipeline using the /orchestrate command defined in commands/orchestrate.md. This enables sequential or parallel execution of multiple ECC agents.

Example workflow:

/orchestrate
  - planner
  - tdd-guide
  - code-reviewer
  - security-reviewer

In this pipeline, the code-reviewer acts as a quality gate between implementation and deep security analysis, ensuring that basic defects are eliminated before expensive security scans run.

Customizing the Review Checklist

Advanced teams can extend the code-reviewer's capabilities by modifying agents/code-reviewer.md. The checklist section supports custom rules, such as prohibiting specific function calls or enforcing internal naming conventions.

After editing the agent definition, commit the changes to your repository. The updated rules automatically apply to all subsequent reviews, allowing the code-reviewer to evolve alongside your project's standards.

Summary

Integrating the code-reviewer agent into a Claude Code development workflow provides automated, immediate feedback on every code change. Key takeaways include:

• The agent is defined in agents/code-reviewer.md and triggered automatically via ECC's hook system after every file edit.
• Reviews analyze git diff output against a security and quality checklist, blocking commits that contain CRITICAL or HIGH severity issues.
• Developers can manually invoke reviews with /code-review, automate them via hooks, or orchestrate them in multi-agent pipelines with /orchestrate.
• The review checklist in agents/code-reviewer.md is fully customizable to enforce project-specific standards.

Frequently Asked Questions

What triggers the code-reviewer agent automatically?

The ECC hook system monitors file write and edit operations within Claude Code. After any file modification, the system automatically invokes the code-reviewer agent as described in AGENTS.md, ensuring immediate feedback without manual commands.

How does the code-reviewer block commits with critical issues?

The commands/code-review.md file implements a gating policy that examines the agent's output. If the review report contains any findings marked as CRITICAL or HIGH severity, the command returns a failure status that prevents the Claude Code toolchain from completing the git commit operation.

Can I run the code-reviewer alongside other ECC agents?

Yes. Use the /orchestrate command defined in commands/orchestrate.md to chain agents sequentially or run them in parallel. A typical integration runs the planner, tdd-guide, code-reviewer, and security-reviewer in sequence to create a comprehensive development pipeline.

Where can I customize the review rules for my project?

The review checklist and agent behavior are defined in agents/code-reviewer.md. You can edit this file to add project-specific rules, such as banning specific imports or enforcing naming conventions. Once committed, these changes automatically apply to all future reviews in the repository.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}