Agents
The Compound Engineering plugin includes 29 specialized agents organized into categories for easier discovery. Each agent is an expert in a specific domain, bringing focused knowledge and best practices to your development workflow.
What Are Agents?
Agents are autonomous AI assistants that can be spawned using the Task tool. They run independently and in parallel, each bringing specialized expertise to analyze, review, or research specific aspects of your codebase.
How to use agents:
Task agent-name: "Description of what you want the agent to do"
Task security-sentinel: "Review this authentication flow"
Task performance-oracle: "Analyze database query performance"
Task kieran-rails-reviewer: "Review Rails conventions"
Review Agents (15)
Review agents provide specialized code review from different perspectives, catching issues that generic linters might miss.
| Agent | Description | When to Use |
|---|
agent-native-reviewer | Verify features are agent-native (action + context parity) | Building AI-first features with proper tool design |
architecture-strategist | Analyze architectural decisions and compliance | Major refactors, system design reviews |
code-simplicity-reviewer | Final pass for simplicity and minimalism | Before merging, to ensure code isn’t over-engineered |
data-integrity-guardian | Database migrations and data integrity | Any schema changes, migration reviews |
data-migration-expert | Validate ID mappings match production, check for swapped values | Data migrations, production data changes |
deployment-verification-agent | Create Go/No-Go deployment checklists for risky data changes | Pre-deployment validation, production changes |
dhh-rails-reviewer | Rails review from DHH’s perspective | Rails codebases following 37signals style |
julik-frontend-races-reviewer | Review JavaScript/Stimulus code for race conditions | Frontend code with async operations |
kieran-rails-reviewer | Rails code review with strict conventions | Rails projects requiring strict adherence to conventions |
kieran-python-reviewer | Python code review with strict conventions | Python projects requiring best practices |
kieran-typescript-reviewer | TypeScript code review with strict conventions | TypeScript/JavaScript projects |
pattern-recognition-specialist | Analyze code for patterns and anti-patterns | Code audits, identifying technical debt |
performance-oracle | Performance analysis and optimization | Performance-critical code, optimization work |
schema-drift-detector | Detect unrelated schema.rb changes in PRs | Rails PRs to catch accidental schema changes |
security-sentinel | Security audits and vulnerability assessments | Any code touching auth, payments, user data |
Review workflow: The /ce:review command automatically runs multiple review agents in parallel and consolidates their feedback. You can also invoke individual review agents directly for focused analysis.
Research Agents (5)
Research agents gather context, documentation, and best practices to inform implementation decisions.
| Agent | Description | When to Use |
|---|
best-practices-researcher | Gather external best practices and examples | Unfamiliar domains, new technologies |
framework-docs-researcher | Research framework documentation and best practices | Framework-specific features, API usage |
git-history-analyzer | Analyze git history and code evolution | Understanding why code exists, finding context |
learnings-researcher | Search institutional learnings for relevant past solutions | Checking if team has solved this before |
repo-research-analyst | Research repository structure and conventions | New codebases, understanding project patterns |
Task repo-research-analyst: "Find existing authentication patterns in this codebase"
Task learnings-researcher: "Search for past solutions related to API rate limiting"
Task framework-docs-researcher: "Research Rails ActiveJob best practices for background processing"
/ce:plan to inform the planning process with real context.
Design Agents (3)
Design agents ensure UI implementations match designs and help iterate on visual design.
| Agent | Description | When to Use |
|---|
design-implementation-reviewer | Verify UI implementations match Figma designs | After implementing UI from designs |
design-iterator | Iteratively refine UI through systematic design iterations | Improving UI/UX through rapid iteration |
figma-design-sync | Synchronize web implementations with Figma designs | Keeping code in sync with design files |
Design agents work best when you have Figma designs to reference. They can help catch visual regressions and ensure pixel-perfect implementations.
Workflow Agents (5)
Workflow agents automate common development tasks and processes.
| Agent | Description | When to Use |
|---|
bug-reproduction-validator | Systematically reproduce and validate bug reports | Before fixing bugs, to ensure you understand them |
every-style-editor | Edit content to conform to Every’s style guide | Content editing for Every publications |
lint | Run linting and code quality checks on Ruby and ERB files | Pre-commit checks, CI/CD pipelines |
pr-comment-resolver | Address PR comments and implement fixes | After receiving PR feedback |
spec-flow-analyzer | Analyze user flows and identify gaps in specifications | Validating feature specs, finding edge cases |
# The /resolve_pr_parallel command uses pr-comment-resolver agent
/resolve_pr_parallel
pr-comment-resolver agents to address each PR comment simultaneously.
Docs Agents (1)
Documentation agents help create and maintain high-quality documentation.
| Agent | Description | When to Use |
|---|
ankane-readme-writer | Create READMEs following Ankane-style template for Ruby gems | Creating new Ruby gem projects |
The Ankane style is known for clear, concise documentation with practical examples. This agent helps maintain that standard across Ruby gem projects.
Agent Locations
Agents are organized in the plugin directory structure:
plugins/compound-engineering/agents/
├── review/ # 15 review agents
│ ├── agent-native-reviewer.md
│ ├── architecture-strategist.md
│ ├── code-simplicity-reviewer.md
│ └── ...
├── research/ # 5 research agents
│ ├── best-practices-researcher.md
│ ├── framework-docs-researcher.md
│ └── ...
├── design/ # 3 design agents
│ ├── design-implementation-reviewer.md
│ ├── design-iterator.md
│ └── figma-design-sync.md
├── workflow/ # 5 workflow agents
│ ├── bug-reproduction-validator.md
│ ├── pr-comment-resolver.md
│ └── ...
└── docs/ # 1 documentation agent
└── ankane-readme-writer.md
Advanced Usage
Running Agents Programmatically
You can spawn agents from within commands or skills:
# In a command or skill
- Task security-sentinel(code_to_review)
- Task performance-oracle(database_queries)
Custom Agent Configuration
Some agents respect project-specific configuration in CLAUDE.md. For example, the review agents can be configured using the setup skill:
# Configure which review agents run for your project
/ce:review --setup
Combining Agents with Skills
Agents often work alongside skills. For example:
- The
dhh-rails-reviewer agent uses knowledge from the dhh-rails-style skill
- The
agent-native-reviewer uses patterns from the agent-native-architecture skill
This combination of agent expertise + skill documentation creates powerful, context-aware reviews.
See Also
- Commands - Slash commands that orchestrate agents
- Skills - Specialized knowledge domains that agents can reference
- Workflow Guide - How agents fit into development workflows