PolicyEngine Claude Agents

PolicyEngine Claude Agents

This repository contains specialized Claude agents for PolicyEngine development across different types of repositories.

Repository Structure

agents/
├── country-models/              # Agents for country packages (policyengine-us, policyengine-uk, etc.)
│   ├── ci-fixer.md              # Fixes CI issues iteratively
│   ├── cross-program-validator.md # Validates program interactions
│   ├── document-collector.md    # Collects authoritative sources
│   ├── documentation-enricher.md # Enriches code with examples
│   ├── edge-case-generator.md   # Generates edge case tests
│   ├── implementation-validator.md # Validates implementations
│   ├── integration-agent.md     # Advanced merge workflows
│   ├── isolation-enforcement.md # Enforces agent isolation
│   ├── isolation-setup.md       # Sets up isolation environment
│   ├── issue-manager.md         # Manages GitHub issues and PRs
│   ├── performance-optimizer.md # Optimizes calculations
│   ├── pr-pusher.md             # Formats and pushes PRs
│   ├── program-reviewer.md      # Reviews regulatory compliance
│   ├── rules-engineer.md        # Implements rules from documentation
│   ├── test-creator.md          # Creates tests from documentation
│   └── workflow.md              # Multi-agent workflow documentation
├── api/                         # Agents for policyengine-api
│   └── api-reviewer.md          # Reviews API implementations
├── app/                         # Agents for policyengine-app and web tools
│   ├── app-reviewer.md          # Reviews React app code
│   ├── seo-meta-checker.md      # Audits SEO meta tags and OG tags
│   ├── seo-crawlability-checker.md # Audits robots.txt, sitemap, routing
│   ├── seo-performance-checker.md  # Audits bundle sizes and performance
│   └── seo-content-checker.md   # Audits semantic HTML and content
├── shared/                      # Shared resources across all repos
│   ├── policyengine-standards.md # Common standards and patterns
│   ├── model-evaluator.md       # Evaluates model outputs
│   └── pr-merge-checklist.md    # PR merge checklist
├── branch-comparator.md         # Compares branches for differences
├── legislation-statute-analyzer.md # Analyzes legislative text
└── reference-validator.md       # Validates parameter references

Usage

Install via Claude Code plugin marketplace:

/plugin marketplace add PolicyEngine/policyengine-claude
/plugin install country-models@policyengine-claude

Or add to your repo's .claude/settings.json for auto-install:

{
  "plugins": {
    "marketplaces": ["PolicyEngine/policyengine-claude"],
    "auto_install": ["country-models@policyengine-claude"]
  }
}

Agent Categories

Country Models (policyengine-us, policyengine-uk, etc.)

These agents support the multi-agent development workflow for implementing tax and benefit rules with proper isolation and verification.

API (policyengine-api)

Agents focused on Flask API development, performance, security, and proper REST practices.

App (policyengine-app)

Agents for React application development, focusing on component quality, performance, and user experience.

Shared

Resources and agents that apply across all PolicyEngine repositories.

Multi-Agent Workflow (Country Models)

For country model development, we use an isolated multi-agent approach:

1. Document Collector gathers authoritative sources
2. Test Creator writes tests (without seeing implementation)
3. Rules Engineer implements rules (without seeing test expectations)
4. Implementation Validator verifies implementation quality and compliance
5. Supervisor orchestrates and ensures quality

See country-models/workflow.md for detailed workflow documentation.

Complete Agent Directory

Country Models Agents (15)

Agent	Description	Primary Use Case
ci-fixer	Runs tests locally, fixes issues iteratively	Automated CI/CD pipeline fixing
cross-program-validator	Validates interactions between benefit programs	Preventing integration issues
document-collector	Gathers authoritative documentation	Research and documentation collection
documentation-enricher	Enriches code with examples and references	Improving code documentation
edge-case-generator	Generates comprehensive edge case tests	Test coverage improvement
implementation-validator	Validates implementations for quality	Code quality assurance
integration-agent	Advanced merge workflows	Branch management
issue-manager	Finds or creates GitHub issues and PRs	Issue tracking and management
performance-optimizer	Optimizes benefit calculations for performance	Performance tuning
pr-pusher	Formats and pushes PRs	PR quality control
program-reviewer	Reviews government program implementations	Regulatory compliance
rules-engineer	Implements government benefit program rules	Policy implementation
test-creator	Creates comprehensive integration tests	Test development
isolation-setup	Sets up git worktrees for isolated development	Development environment
isolation-enforcement	Enforces test/implementation isolation	Multi-agent workflow

Root-Level Agents (3)

Agent	Description	Primary Use Case
branch-comparator	Compares branches for differences	Identifying changes between development branches
legislation-statute-analyzer	Analyzes legislative text and identifies statutes	Legal document analysis
reference-validator	Validates that all parameters have proper references	Documentation validation

API-Specific Agents

Agent	Description	Location
api-reviewer	Reviews API implementations for REST best practices	api/

App-Specific Agents

Agent	Description	Location
app-reviewer	Reviews React app code for quality and performance	app/
seo-meta-checker	Audits meta tags, OG tags, Twitter cards, canonical URLs	app/
seo-crawlability-checker	Audits robots.txt, sitemap, routing, SSR, hosting config	app/
seo-performance-checker	Audits bundle sizes, code splitting, fonts, images	app/
seo-content-checker	Audits heading hierarchy, semantic HTML, accessibility	app/

Shared Resources

Resource	Description	Purpose
policyengine-standards	Common standards and patterns	Code consistency
model-evaluator	Evaluates model outputs	Quality assurance
pr-merge-checklist	PR merge checklist	Quality control

Key Principles

1. Source Authority: Statutes > Regulations > Websites
2. Isolation: Tests and implementation developed separately
3. Vectorization: No if-elif-else with household data
4. Documentation: Every value traces to primary source
5. Testing: Document calculations with regulation references

Total Agent Count

• Country Models: 15 agents
• Root-Level: 3 agents
• API: 1 agent
• App: 5 agents
• Total: 24 agents