Code Reviewer

Code Reviewer — Automated Code Review Pipeline

An agent team systematically reviews code across style, security, performance, and architecture.

Execution Mode

Agent Team — 5 members communicate directly via SendMessage and cross-validate each other's work.

Agent Composition

Workflow

Phase 1: Preparation (Performed directly by the orchestrator)

1. Extract from user input:
   • Target Code: File paths, PR number, diff, directory
   • Language/Framework: Auto-detect or user-specified
   • Review Scope (optional): If only specific domains were requested
   • Context (optional): PR description, related issues, change rationale
   • Style Guide (optional): Team-specific conventions
2. Create the _workspace/ directory at the project root
3. Organize the input and save to _workspace/00_input.md
4. Identify the target code and determine the review scope
5. If existing files are provided, copy them to _workspace/ and skip the corresponding phase
6. Determine the execution mode based on the scope of the request

Phase 2: Team Assembly and Execution

Tasks 1a-1d (all 4 domain reviews) are all executed in parallel.

Inter-team communication flow:

• style-inspector -> Delivers sensitive info in comments to security-analyst, complex function lists to performance-analyst
• security-analyst -> Delivers security measure performance impact to performance-analyst, authentication architecture to architecture-reviewer
• performance-analyst -> Delivers structural bottlenecks to architecture-reviewer
• review-synthesizer integrates all reviews. Requests additional analysis from relevant analysts when cross-domain conflicts are found

Phase 3: Integration and Final Artifacts

Organize the final artifacts based on the comprehensive report:

1. Verify all reviews in _workspace/
2. Determine the final verdict (Approve/Request Changes/Reject)
3. Report the final summary to the user

Mode by Task Scale

PR Review: When a PR number is provided, extract the diff and focus review on changed code. Reference full file context but concentrate the review on the diff.

Data Transfer Protocol

File naming convention: {order}_{agent}_{artifact}.{extension}

Error Handling

Test Scenarios

Normal Flow

Prompt: "Do a full code review of this Python Flask project" + code files/directory Expected Result:

• Style: PEP 8 standards, naming/formatting/readability checks, Black/flake8 config suggestions
• Security: SQL injection, XSS, hardcoded secrets, dependency CVE checks
• Performance: Query optimization, N+1, memory usage, caching opportunities
• Architecture: MVC pattern compliance, SOLID, dependency analysis
• Comprehensive: Unified priorities, final verdict, action items

Existing File Flow

Prompt: "Review only the security of this PR" + PR diff Expected Result:

• Security mode: deploy security-analyst + review-synthesizer
• Diff-focused review, full file context for reference
• Skip style-inspector, performance-analyst, architecture-reviewer

Error Flow

Prompt: "Look at this code" + single file (under 100 lines) Expected Result:

• Small codebase -> Architecture review shifts to function separation/module design perspective
• Run in full review mode, but each domain adjusts to code scale
• Comprehensive report notes "single file review, architecture assessment limited"

Agent Extension Skills

Extension skills that enhance each agent's domain expertise: