How to Implement Code Review Automation in ECC: A Complete Guide

Implement code review automation in ECC by leveraging the built-in code-reviewer agent, which can be triggered via the /code-review command, integrated into orchestrated workflows, or embedded into CI pipelines and Git hooks to automatically analyze code changes for quality, security, and maintainability.

The Everything Claude Code (ECC) platform from the affaan-m/ECC repository provides a comprehensive, LLM-driven solution to implement code review automation directly into your development workflow. By utilizing the code-reviewer agent and its associated commands, you can enforce consistent quality standards across local development, continuous integration, and pull request processes without manual intervention.

Core Components of ECC Code Review Automation

The code-reviewer Agent

The heart of the system lives in agents/code-reviewer.md, which defines the LLM-driven reviewer. This file specifies the review checklist, confidence thresholds, and structured output format. The agent implements severity bands ranging from CRITICAL to LOW, ensuring comprehensive coverage of potential issues.

The /code-review Command

Located in docs/zh-CN/commands/code-review.md, this command serves as the primary user-facing entry point. When invoked, it launches the reviewer agent on the current diff, making it ideal for ad-hoc reviews during development sessions.

Orchestration Engine Integration

The orchestrator, documented in docs/zh-CN/commands/orchestrate.md, automatically inserts the code-reviewer into workflows for impl, refactor, or migration steps. When you run /orchestrate feature "...", the engine builds a chain such as planner → tdd-guide → code-reviewer, ensuring reviews happen automatically after implementation.

How the Code Review Process Works

The automation follows a four-step pipeline:

1. Gather Change Context – Executes git diff --staged or falls back to recent commits to isolate exact changes.
2. Read Affected Files – Uses the Read tool to load full source files, imports, and related test files for broader context.
3. Apply the Review Checklist – Processes the checklist from agents/code-reviewer.md, filtering by confidence levels.
4. Produce Structured Output – Generates findings grouped by severity with a final verdict of APPROVE, WARNING, or BLOCK.

Confidence-Based Filtering

According to lines 31-34 of the agent definition in agents/code-reviewer.md, the system filters noise by only reporting issues with greater than 80% confidence. This threshold prevents low-probability suggestions from cluttering results while maintaining high signal-to-noise ratios for critical findings.

Integration Methods for Automation

Manual Ad-hoc Reviews

Trigger immediate reviews during Claude Code sessions:

/code-review

This command analyzes the current diff and returns a formatted report with severity counts and a verdict.

Orchestrated Workflows

For structured development, use workflow commands:

/orchestrate feature "Add JWT authentication"

The orchestrator automatically chains agents: planner → tdd-guide → code-reviewer → security-reviewer. The review runs immediately after implementation completes, blocking progression if issues are found.

CI Pipeline Integration

Automate reviews in CI using the headless mode. The test file tests/ci/code-reviewer-false-positive-guard.test.js validates that the reviewer maintains consistent formatting and guard thresholds.

Example GitHub Actions workflow:

# In .github/workflows/ci.yml

- name: Run ECC code review
  run: |
    npx claude-code review \
      --repo $GITHUB_WORKSPACE \
      --diff $(git diff HEAD~1) \
      --output review.json
    cat review.json

If the JSON output contains BLOCK, the workflow exits with a non-zero status, failing the pull request.

Git Pre-commit Hooks

Prevent low-quality commits at the source by creating .git/hooks/pre-commit:

#!/usr/bin/env bash

# Run ECC's code reviewer on staged changes

if ! claude-code review --staged; then
  echo "Code review failed – fix issues before committing."
  exit 1
fi

Make the file executable (chmod +x .git/hooks/pre-commit). Now every git commit triggers an automatic review; commits proceed only if the reviewer returns APPROVE.

Implementation Examples

Custom Orchestration with Explicit Ordering

Control review timing by manually composing agent chains:

/orchestrate custom "architect,tdd-guide,code-reviewer,security-reviewer" "Redesign caching layer"

This places the code reviewer before security checks or after specific architecture decisions, giving you fine-grained control over the validation sequence.

Verdict-Based Workflow Control

The structured output enables automated decision-making:

## Review Summary

| Severity | Count | Status |
|----------|-------|--------|
| CRITICAL | 0     | pass   |
| HIGH     | 1     | warn   |
| MEDIUM   | 2     | info   |
| LOW      | 0     | note   |
Verdict: WARNING — resolve the HIGH issue before merging.

Parse this output in CI scripts to conditionally block merges or notify reviewers.

Key Configuration Files

File	Purpose
agents/code-reviewer.md	Full agent definition including checklist and output format
docs/zh-CN/commands/code-review.md	Documentation for the /code-review command
docs/zh-CN/commands/orchestrate.md	Orchestrator integration patterns
tests/ci/code-reviewer-false-positive-guard.test.js	CI validation ensuring reviewer headings and confidence guards remain intact
docs/COMMAND-AGENT-MAP.md	Maps commands to agents; identifies code-reviewer as the default for "no-tag" steps
tests/ci/validators.test.js	Test harness validating the presence and shape of the reviewer file

Summary

• Implement code review automation in ECC using the code-reviewer agent defined in agents/code-reviewer.md, triggered by the /code-review command or orchestrated workflows.
• Apply 80% confidence filtering to eliminate noise while catching critical issues.
• Integrate into CI pipelines using headless CLI commands to enforce quality gates on every pull request.
• Use Git pre-commit hooks to block low-quality code before it enters the repository.
• Leverage the orchestration engine to automatically insert reviews into feature development, refactoring, and migration workflows.

Frequently Asked Questions

What is the minimum confidence threshold for ECC code review reports?

The code-reviewer agent only reports issues with greater than 80% confidence, as specified in lines 31-34 of agents/code-reviewer.md. This filtering mechanism prevents false positives and ensures developers receive high-signal feedback.

How do I integrate ECC code review into my CI pipeline?

Configure your CI workflow to run npx claude-code review with the --repo and --diff flags, outputting to JSON. The process validates the same checklist used in interactive mode. If the output contains a BLOCK verdict or CRITICAL severity issues, exit with a non-zero status to fail the build, as demonstrated in tests/ci/code-reviewer-false-positive-guard.test.js.

Can I run ECC code review automatically before every commit?

Yes. Create an executable .git/hooks/pre-commit script that runs claude-code review --staged. If the reviewer returns anything other than APPROVE, the hook exits with code 1, preventing the commit from completing until issues are resolved.

Where is the code-reviewer agent behavior defined?

The complete behavior, including the review checklist, severity bands (CRITICAL → LOW), and output formatting rules, is defined in agents/code-reviewer.md at the repository root. This file serves as the source of truth for how the LLM analyzes code quality and security patterns.

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