review-agent

name	review-agent
description	Read code changes with adversarial intent to find bugs, security holes, logic errors, and performance traps. Use when reviewing PRs, auditing refactoring for regressions, or running pre-deploy safety checks.
domain	agents
tags	["agent","ai-agent","automation","orchestration","review"]

Review Agent

When to Use

Trigger phrases:

• "review agent"

• "Reviewing pull requests before merge"

• "Auditing code changes for security issues"

• "Validating refactoring has not introduced regressions"

• Reviewing pull requests before merge

• Auditing code changes for security issues

• Validating refactoring has not introduced regressions

• Checking that new features handle edge cases

• Reviewing third-party library integrations

• Pre-deploy safety checks

• Post-mortem analysis of production incidents

When NOT to Use

• When the task is simple enough for a single command
• When real-time human judgment is required
• When the agent lacks access to required tools or data

Overview

Review Agent is an AI agent skill for agent orchestration. It enables autonomous execution of complex tasks with minimal human intervention.

Capabilities

• Autonomous operation — Execute multi-step review agent workflows independently
• Context awareness — Adapt behavior based on current state and history
• Error recovery — Handle failures gracefully with retry and fallback logic
• Integration — Connect with external tools and services as needed

Workflow

# Example: Agent orchestration
from dataclasses import dataclass

@dataclass
class Task:
    name: str
    priority: int
    assigned_agent: str

def orchestrate(tasks: list[Task]) -> dict:
    results = {}
    for task in sorted(tasks, key=lambda t: t.priority):
        results[task.name] = execute(task)
    return results

1. Initialize — Set up the agent context and load required resources
2. Plan — Break down the task into executable steps
3. Execute — Run each step, monitoring for errors and adapting as needed
4. Verify — Validate results against acceptance criteria
5. Report — Summarize outcomes and suggest next steps

Configuration

• Define task objectives and constraints clearly
• Set appropriate timeout and retry limits
• Configure tool access and permissions
• Enable logging for debugging and audit

Anti-Rationalization

Rationalization	Reality
"I will just do it manually"	Agents automate repetitive tasks — manual work does not scale
"The agent will figure it out"	Without clear instructions, agents hallucinate. Give explicit context.
"One agent is enough"	Complex tasks benefit from specialized agents working in parallel

Process

1. Prepare — Gather requirements, verify prerequisites, set up environment
2. Execute — Run review agent workflow with configured parameters
3. Verify — Validate output meets requirements, document results

Verification

• All steps executed successfully
• Results validated against acceptance criteria
• Error handling tested with edge cases
• Documentation updated with findings