Specialist Agents
Relevant source files
The following files were used as context for generating this wiki page:
Specialist Agents within the SuperGemini Framework are domain-specific AI experts designed to transform the general-purpose Gemini CLI into a structured development platform. These agents operate through a "Persona Mode" where the AI embodies a specific role—such as a security engineer, frontend architect, or technical writer—rather than spawning separate processes. This approach leverages behavioral instruction injection via Markdown configuration files to modify AI behavior dynamically based on the task at hand.
Sources: Docs/User-Guide/agents.md:16-24, CLAUDE.md:143-145
The system employs intelligent routing to coordinate 13 specialized agents through established communication protocols and decision hierarchies. Agents are not manually selected by the user; instead, they utilize an auto-activation mechanism triggered by specific keywords, file types, and task complexity scores. This allows for collaborative synthesis where multiple specialists can work together on multi-domain tasks, ensuring high-quality, professional software development workflows.
Sources: CLAUDE.md:83-85, Docs/User-Guide/agents.md:21-24, Docs/Developer-Guide/README.md:32-34
Agent Architecture and Activation
The Specialist Agent system is built on a configuration distribution model. Each agent is defined by a Markdown file located in SuperGemini/Agents/. These files contain metadata (name, description, category), trigger keywords, behavioral mindsets, and key actions that the AI must follow when the persona is active.
Sources: CLAUDE.md:22-26, SuperGemini/Agents/technical-writer.md:1-24
Activation Logic
Agents engagement is governed by a priority hierarchy to ensure the most relevant expertise is applied to a user request.
The diagram shows the decision tree for agent selection, prioritizing keywords and file types before assessing complexity for multi-agent coordination.
Sources: Docs/User-Guide/agents.md:37-60
Selection Rules Table
The following table outlines the criteria used by the framework to select and resolve agent conflicts.
| Rule Type | Description |
|---|---|
| Keywords | Direct domain terminology triggers primary agents (e.g., "security" triggers security-engineer). |
| File Types | Extensions activate language specialists (e.g., .py triggers python-expert). |
| Complexity | Tasks with >5 components or multi-step requirements engage system-architect oversight. |
| Conflict Resolution | Unclear context triggers requirements-analyst; quality concerns include QA/Security agents. |
Sources: Docs/User-Guide/agents.md:37-45, Docs/User-Guide/agents.md:99-105
Agent Categories and Domains
The framework organizes its 13 agents into functional groups that cover the entire software development lifecycle (SDLC).
Architecture and Design Agents
These agents focus on high-level system structure, backend reliability, and infrastructure automation.
- system-architect: Handles large-scale distributed design and scalability.
- backend-architect: Focuses on API reliability, data integrity, and server-side logic.
- devops-architect: Manages CI/CD pipelines, Docker, and Kubernetes orchestration.
Sources: Docs/User-Guide/agents.md:91-160
Quality and Analysis Agents
Focused on the "Quality Gate" philosophy, these agents ensure the robustness and security of the codebase.
- security-engineer: Conducts threat modeling and vulnerability prevention.
- performance-engineer: Identifies bottlenecks and optimizes resource efficiency.
- quality-engineer: Designs testing strategies and automated validation frameworks.
- root-cause-analyst: Investigates system failures through systematic evidence-based analysis.
Sources: Docs/User-Guide/agents.md:183-260
Communication and Specialized Agents
These agents bridge the gap between technical implementation and user/developer understanding.
- technical-writer: Specializes in API documentation, user guides, and information architecture.
- learning-guide: Designs educational content and progressive learning paths for complex concepts.
- python-expert: Provides deep expertise in modern Python frameworks like FastAPI and Django.
Sources: SuperGemini/Agents/technical-writer.md:1-12, Docs/User-Guide/agents.md:330-380
Coordination and Tool Integration
Specialist Agents do not work in isolation; they are integrated with the Model Context Protocol (MCP) and SuperGemini slash commands.
Sequence of Multi-Agent Interaction
When a complex command like /sg:implement is used, the framework orchestrates a sequence of agent activations and tool uses.
This sequence illustrates how the orchestrator coordinates a primary architect with security and quality specialists, supported by MCP tool data.
Sources: Docs/User-Guide/agents.md:73-82, CLAUDE.md:392-410, SuperGemini/Commands/index.toml:13-30
Core Command-Agent Mappings
Different slash commands are weighted to trigger specific primary agents.
| Command | Primary Agents | Supporting Agents |
|---|---|---|
/sg:implement | Frontend/Backend Architects | Security Engineer, Quality Engineer |
/sg:analyze | Quality Engineer, Security Engineer | Performance Engineer, Root-Cause-Analyst |
/sg:design | System Architect | Domain Architects, Requirements Analyst |
/sg:document | Technical Writer | Domain Specialists, Learning Guide |
Sources: Docs/User-Guide/agents.md:520-535, SuperGemini/Commands/index.toml:7-11
Implementation Details
Agents are configured via YAML frontmatter and Markdown sections. For example, the Technical Writer agent configuration includes specific focus areas and outputs.
---
name: technical-writer
description: Create clear, comprehensive technical documentation
category: communication
---
# Technical Writer
## Triggers
- API documentation and technical specification creation requests
- User guide development needs
Sources: SuperGemini/Agents/technical-writer.md:1-10
The Framework's index command demonstrates how these agents are grouped at the configuration level to handle complex tasks like generating a project knowledge base.
name = "index"
description = "Generate comprehensive project documentation"
agents = ["system-architect", "technical-writer", "quality-engineer"]
mcp_servers = ["sequential", "context7", "superagent:gemini"]
Sources: SuperGemini/Commands/index.toml:1-5
Specialist Agents are the cornerstone of the SuperGemini Framework's ability to provide professional-grade software development assistance. By partitioning AI intelligence into 13 specialized domains and employing an automated orchestration layer, the system ensures that every development task—from architecture design to security auditing—is handled with appropriate expertise and validated through specialized quality gates.