Agent Reference

Complete reference for all 203 local specialized AI agents organized by category with model assignments. Use it as a repeatable review, validation or hardening pass.

Editorial note: curated source snapshot published by Collider.club under the MIT License. Source attribution is preserved in the front matter.

Source snapshot

Agent Reference

Complete reference for all 203 local specialized AI agents organized by category with model assignments.

Agent Categories

Architecture & System Design

Core Architecture

UI/UX & Mobile

Programming Languages

Systems & Low-Level

Web & Application

Enterprise & JVM

Specialized Platforms

Infrastructure & Operations

DevOps & Deployment

Database Management

Incident Response & Network

Project Management

Quality Assurance & Security

Code Quality & Review

Testing & Debugging

Performance & Observability

Data & AI

Data Engineering & Analytics

Machine Learning & AI

Documentation & Technical Writing

Business & Operations

Business Analysis & Finance

Marketing & Sales

Support & Legal

SEO & Content Optimization

Specialized Domains

Model Configuration

Agents are assigned to specific Claude models based on task complexity and computational requirements.

Model Distribution Summary

Model Selection Criteria

Haiku - Fast Execution & Deterministic Tasks

Use when:

• Generating code from well-defined specifications
• Creating tests following established patterns
• Writing documentation with clear templates
• Executing infrastructure operations
• Performing database query optimization
• Handling customer support responses
• Processing SEO optimization tasks
• Managing deployment pipelines

Sonnet - Complex Reasoning & Architecture

Use when:

• Designing system architecture
• Making technology selection decisions
• Performing security audits
• Reviewing code for architectural patterns
• Creating complex AI/ML pipelines
• Providing language-specific expertise
• Orchestrating multi-agent workflows
• Handling business-critical legal/HR matters

Fable - Longest-Horizon Autonomous Work

Claude Fable 5 (model: fable) is the tier above Opus. It is opt-in in Claude Code (v2.1.170+, never the default) and carries roughly 2.6× the effective cost of Opus ($10/$50 per MTok plus a ~30% heavier tokenizer), so reserve it for agents where Opus demonstrably needs multiple attempts.

Use when:

• Running multi-hour autonomous sessions (large codebase migrations, overnight refactors)
• Executing end-to-end work from a single well-specified goal
• Coordinating long-lived parallel sub-agent workstreams

Avoid for:

• Security-analysis agents — Fable 5's cyber/bio safety classifiers fall back to Opus 4.8 on that content anyway
• Anything Sonnet or Opus already completes in one pass

Other harnesses map fable to their top available model (see authoring.md). Fable-tier agent bodies should state goals and constraints rather than step-by-step scaffolding, and must not ask the model to echo its internal reasoning.

Hybrid Orchestration Patterns

The plugin ecosystem leverages Sonnet + Haiku orchestration for optimal performance and cost efficiency:

Pattern 1: Planning → Execution

Sonnet: backend-architect (design API architecture)
  ↓
Haiku: Generate API endpoints following spec
  ↓
Haiku: test-automator (generate comprehensive tests)
  ↓
Sonnet: code-reviewer (architectural review)

Pattern 2: Reasoning → Action (Incident Response)

Sonnet: incident-responder (diagnose issue, create strategy)
  ↓
Haiku: devops-troubleshooter (execute fixes)
  ↓
Haiku: deployment-engineer (deploy hotfix)
  ↓
Haiku: Implement monitoring alerts

Pattern 3: Complex → Simple (Database Design)

Sonnet: database-architect (schema design, technology selection)
  ↓
Haiku: sql-pro (generate migration scripts)
  ↓
Haiku: database-admin (execute migrations)
  ↓
Haiku: database-optimizer (tune query performance)

Pattern 4: Multi-Agent Workflows

Full-Stack Feature Development:
Sonnet: backend-architect + frontend-developer (design components)
  ↓
Haiku: Generate code following designs
  ↓
Haiku: test-automator (unit + integration tests)
  ↓
Sonnet: security-auditor (security review)
  ↓
Haiku: deployment-engineer (CI/CD setup)
  ↓
Haiku: Setup observability stack

Agent Invocation

Natural Language

Agents can be invoked through natural language when you need Claude to reason about which specialist to use:

"Use backend-architect to design the authentication API"
"Have security-auditor scan for OWASP vulnerabilities"
"Get performance-engineer to optimize this database query"

Slash Commands

Many agents are accessible through plugin slash commands for direct invocation:

/backend-development:feature-development user authentication
/security-scanning:security-sast
/incident-response:smart-fix "memory leak in payment service"

Contributing

To add a new agent:

1. Create plugins/{plugin-name}/agents/{agent-name}.md
2. Add frontmatter with name, description, and model assignment
3. Write comprehensive system prompt
4. Update plugin definition in .claude-plugin/marketplace.json

See Contributing Guide for details.

About Collider.club

This card belongs to the curated knowledge base of Collider.club — a closed business club for entrepreneurs, engineers, investors and domain experts building projects for international markets. Members work across DeFi, AI/ML, FinTech, Web3, banking, hardware and venture capital, and the club runs closed sessions on high-margin niches with anonymous speakers.

• Club: https://collider.club
• Collection: Collider.club curated card library (mdrss-card/v2)
• Maintainer: Collider.club editorial team

License

MIT License — Copyright (c) 2026 Collider.club. Full text: LICENSE · https://opensource.org/licenses/MIT