Agents Reference
Agents Reference
Section titled “Agents Reference”PopKit provides 23 specialized AI agents across 4 plugins. Agents are organized into tiers based on activation frequency.
Agent Tiers
Section titled “Agent Tiers”| Tier | Description | Activation |
|---|---|---|
| Tier 1: Always Active | Core agents for common tasks | Automatically available in every session |
| Tier 2: On-Demand | Specialized agents for specific needs | Activated when relevant tasks are detected |
| Feature Workflow | Development phase agents | Used during feature development phases |
popkit-core (9 agents)
Section titled “popkit-core (9 agents)”Foundation agents for accessibility, API design, documentation, and orchestration.
Tier 1: Always Active
Section titled “Tier 1: Always Active”accessibility-guardian
Section titled “accessibility-guardian”Description: Ensures web applications meet WCAG 2.1 AA/AAA compliance. Use when auditing accessibility, fixing a11y violations, or implementing inclusive design patterns.
Tools: Read, Grep, Glob, WebFetch
api-designer
Section titled “api-designer”Description: Expert in RESTful and GraphQL API design patterns. Use when designing new APIs, restructuring existing endpoints, or when you need guidance on API best practices, versioning, and integration patterns.
Tools: Read, Write, Edit, MultiEdit, Grep, WebFetch
documentation-maintainer
Section titled “documentation-maintainer”Description: Keeps documentation synchronized with codebase changes. Use after major feature updates, API changes, or when documentation drift is detected.
Tools: Read, Write, Edit, MultiEdit, Grep, Glob, WebFetch
migration-specialist
Section titled “migration-specialist”Description: Expert in planning and executing complex system migrations including database migrations, API version transitions, framework upgrades, and cloud migrations. Minimizes downtime and ensures data integrity.
Tools: Read, Write, Edit, MultiEdit, Grep, Glob, Bash (database migration tools, version management, testing, backup verification)
Tier 2: On-Demand
Section titled “Tier 2: On-Demand”bundle-analyzer
Section titled “bundle-analyzer”Description: Analyzes and optimizes JavaScript bundle sizes for web applications. Use for identifying bloated dependencies, implementing code splitting, and reducing bundle size.
Tools: Read, Write, Edit, MultiEdit, Grep, Glob, Bash (webpack-bundle-analyzer, vite-bundle-visualizer, source-map-explorer, build tools, dependency analysis)
dead-code-eliminator
Section titled “dead-code-eliminator”Description: Intelligent dead code detection and elimination using advanced static analysis, dependency tracking, and safe removal strategies. Use for codebase cleanup, bundle size optimization, and maintainability improvement.
Tools: Read, Write, Edit, MultiEdit, Grep, Glob, Bash (ts-unused-exports, unimported, depcheck, knip, testing, type checking, linting)
feature-prioritizer
Section titled “feature-prioritizer”Description: Strategic backlog management and feature prioritization specialist. Use when making product roadmap decisions, prioritizing features, or managing development backlogs.
Tools: Read, Grep, Glob, Write, WebFetch
meta-agent
Section titled “meta-agent”Description: Generates new, complete Claude Code agent configuration files from user descriptions. Use proactively when creating custom agents for specific project needs.
Tools: Write, WebFetch, MultiEdit, Read, Grep
power-coordinator
Section titled “power-coordinator”Description: Orchestrates multi-agent collaboration in Power Mode. Use when coordinating parallel agents working on complex tasks via Redis pub/sub mesh network.
Tools: Read, Write, Bash (redis-cli, system monitoring), Task, TodoWrite
popkit-dev (7 agents)
Section titled “popkit-dev (7 agents)”Development workflow agents for code exploration, architecture, review, and conflict resolution.
Feature Workflow
Section titled “Feature Workflow”code-architect
Section titled “code-architect”Description: Designs feature architectures and implementation blueprints based on codebase patterns. Use during architecture phase when multiple implementation approaches exist and trade-offs need evaluation.
Tools: Read, Grep, Glob, Write
code-explorer
Section titled “code-explorer”Description: Deeply analyzes existing codebase features by tracing execution paths, data flow, and dependencies. Use during exploration phase of feature development or when understanding unfamiliar code.
Tools: Read, Grep, Glob
Tier 1: Always Active
Section titled “Tier 1: Always Active”code-reviewer
Section titled “code-reviewer”Description: Performs comprehensive code reviews focusing on TypeScript, React, and Node.js best practices. Use after implementing significant features or when code quality assessment is needed.
Tools: Read, Grep, Glob, Edit
refactoring-expert
Section titled “refactoring-expert”Description: Code restructuring specialist focused on improving quality, maintainability, and performance without changing external behavior. Use for code smell detection, design pattern application, and systematic codebase improvements.
Tools: Read, Edit, MultiEdit, Grep, Glob, Bash (testing, linting, type checking)
Tier 2: On-Demand
Section titled “Tier 2: On-Demand”merge-conflict-resolver
Section titled “merge-conflict-resolver”Description: AI-powered merge conflict resolution with complexity-based prioritization and architectural intelligence. Detects conflicts, analyzes complexity, prioritizes resolution order, proposes intelligent fixes with reasoning, validates with tests.
Triggers: “merge conflict”, “resolve conflicts”, “fix merge issues”, “git conflicts”
Tools: Read, Write, Edit, Bash (git operations, testing), Task, TodoWrite, AskUserQuestion
prd-parser
Section titled “prd-parser”Description: Transforms Product Requirements Documents into structured, actionable tasks with automatic complexity analysis and technology research. Parses markdown PRDs, extracts features, scores complexity, and generates implementation roadmap.
Triggers: “parse this PRD”, “analyze requirements document”, “break down this spec”, “turn this into tasks”
Tools: Read, Write, Grep, Glob, TodoWrite, WebSearch, Bash (gh, python)
rapid-prototyper
Section titled “rapid-prototyper”Description: Fast MVP development specialist for quick proof-of-concept implementations. Use when building prototypes, validating ideas, or creating minimal viable features quickly.
Tools: Read, Write, Edit, MultiEdit, Grep, Glob, Bash (npm, dev servers, testing, build), WebFetch
popkit-ops (7 agents)
Section titled “popkit-ops (7 agents)”Operations agents for debugging, performance, security, testing, and deployment.
Tier 1: Always Active
Section titled “Tier 1: Always Active”bug-whisperer
Section titled “bug-whisperer”Description: Expert debugging specialist for complex issues. Use when facing hard-to-reproduce bugs, performance anomalies, or mysterious system behaviors that require deep investigation and systematic troubleshooting.
Tools: Read, Grep, Edit, MultiEdit, Bash (debugging, testing, log inspection, git operations), WebFetch
performance-optimizer
Section titled “performance-optimizer”Description: Elite performance engineering specialist that analyzes, diagnoses, and optimizes web application performance across all metrics. Use for performance audits, bottleneck identification, and optimization strategies.
Tools: Read, Grep, Glob, Bash (performance testing, build analysis, profiling, load testing, lighthouse), WebFetch
security-auditor
Section titled “security-auditor”Description: Comprehensive security specialist for vulnerability assessment, threat analysis, and defensive security implementation. Use when auditing code, analyzing security risks, or implementing security measures.
Tools: Read, Grep, Glob, Bash (security scanning, dependency audits, linting), WebFetch
test-writer-fixer
Section titled “test-writer-fixer”Description: Comprehensive testing specialist for writing, fixing, and optimizing test suites. Use when implementing tests, debugging test failures, or improving test coverage.
Tools: Read, Write, Edit, MultiEdit, Grep, Glob, Bash (testing frameworks, coverage analysis)
Hooks: PostToolUse auto-run-tests
Tier 2: On-Demand
Section titled “Tier 2: On-Demand”deployment-validator
Section titled “deployment-validator”Description: Ensures safe, reliable deployments through comprehensive validation and verification. Use for pre-deployment checks, smoke testing, and deployment verification.
Tools: Read, Grep, Write, Bash (build, CI/CD workflows, container inspection, Kubernetes, health checks), WebFetch, Task
rollback-specialist
Section titled “rollback-specialist”Description: Expert in rapid recovery procedures and safe rollback operations. Use when deployments fail, production issues arise, or emergency recovery is needed.
Tools: Read, Write, Edit, Bash (git rollback, CI/CD, container management, Kubernetes, health verification), WebFetch, Task
devops-automator
Section titled “devops-automator”Description: Expert in CI/CD pipeline configuration, Docker/container orchestration, infrastructure as code, monitoring setup, environment management, and build optimization. Use for DevOps workflow automation, pipeline design, and infrastructure provisioning.
Tools: Read, Write, Edit, Grep, Glob, Bash (Docker, Kubernetes, Terraform, GitHub CLI, build tools, platform CLIs, health checks), Task
popkit-research (1 agent)
Section titled “popkit-research (1 agent)”Knowledge management agent for codebase analysis and agent discovery.
Tier 2: On-Demand
Section titled “Tier 2: On-Demand”researcher
Section titled “researcher”Description: Meta-researcher that analyzes codebases to identify beneficial agents and development opportunities. Use when discovering what agents would be most helpful for a project or expanding the agent ecosystem.
Tools: Read, Grep, Glob, WebFetch, WebSearch, Task
Agent Summary
Section titled “Agent Summary”| Plugin | Tier 1 | Tier 2 | Feature Workflow | Total |
|---|---|---|---|---|
| popkit-core | 4 | 5 | - | 9 |
| popkit-dev | 2 | 3 | 2 | 7 |
| popkit-ops | 4 | 3 | - | 7 |
| popkit-research | - | 1 | - | 1 |
| Total | 10 | 12 | 2 | 24 |
Agent Memory
Section titled “Agent Memory”Agents support persistent memory through expertise files stored in .claude/expertise/<agent-id>/. Memory includes:
- Patterns: Recurring code patterns the agent has observed
- Preferences: Project-specific conventions and style guidelines
- Common Issues: Frequently encountered problems and their solutions
- Project Context: Framework, architecture, and dependency information
Memory is loaded automatically when an agent starts and updated after significant interactions.
Memory Scopes
Section titled “Memory Scopes”Agents can declare memory persistence scope in their frontmatter:
| Scope | Behavior | Use Case |
|---|---|---|
memory: user | Persists across all projects | General preferences, coding style |
memory: project | Persists within this project | Project-specific patterns |
memory: local | Persists within this directory | Directory-specific context |
Example in agent frontmatter:
memory: projectAgent Isolation
Section titled “Agent Isolation”For parallel agent workflows, agents can run in isolated worktrees:
isolation: worktreeThis prevents file conflicts when multiple agents work simultaneously. The worktree is automatically cleaned up when the agent completes.
Next Steps
Section titled “Next Steps”- Learn about Skills Reference
- Explore Commands Reference
- Read the Getting Started Guide