Agent Prompts

A comprehensive collection of AI agent prompts for software engineering workflows.

Quick Start

git clone https://github.com/mccode415/agent.git

Agents Overview

Core Workflow

Agent	Purpose
staff-engineer	Full-lifecycle engineering (v2 with triage, rollback, incremental commits)
agent-orchestrator	Multi-agent workflow coordinator
handoff-protocol	Agent-to-agent communication standard
deep-research	Codebase and external research before planning
system-architect	Impact analysis and architecture decisions

Validation & Quality

Agent	Purpose
security-fortress	Comprehensive security analysis (OWASP, infra, financial)
security-reviewer	Targeted code security review
change-validator-linter	Code quality and linting
change-verifier	Design alignment and use case verification
performance-analyzer	Performance bottleneck detection
dependency-auditor	Security, updates, and license compliance

Development

Agent	Purpose
test-generator	Comprehensive test generation
docs-generator	API and code documentation
api-designer	REST API design and review
refactor-assistant	Code improvement and cleanup

Exploration

Agent	Purpose
codebase-explorer	Map unfamiliar codebases
plan-visualizer	Create visual plan summaries

Domain Specialists

Agent	Purpose
electron-specialist	Electron desktop app development
llm-specialist	LLM integration, prompts, production patterns
rag-specialist	Retrieval-augmented generation systems
database-specialist	Schema design, queries, migrations, optimization
frontend-specialist	React, CSS, performance, accessibility
devops-specialist	CI/CD, Docker, Kubernetes, infrastructure
api-integration-specialist	Third-party APIs, OAuth, webhooks
realtime-specialist	WebSockets, SSE, presence systems
search-specialist	Full-text, vector, and hybrid search
quant-trading-engineer	Trading systems and backtesting

Skills (Commands)

Quick commands invoked with /command syntax.

Skill	Trigger	Purpose
commit	/commit	Create well-formatted git commit
review-pr	/review-pr	Review a pull request
create-pr	/create-pr	Create a pull request
test	/test	Run and analyze tests
lint	/lint	Run linters and fix issues
explain	/explain	Explain code or concepts
debug	/debug	Debug an issue
refactor	/refactor	Refactor code

Orchestration Workflows

Trigger multi-agent workflows with keywords:

Keyword	Agents Run
full-review	security-fortress + change-validator-linter + change-verifier + performance-analyzer
pre-deploy	security-fortress + dependency-auditor + system-architect + test-generator
new-feature	codebase-explorer + system-architect + api-designer
security-audit	security-fortress + security-reviewer + dependency-auditor
code-quality	change-validator-linter + change-verifier + refactor-assistant + test-generator
trading-review	quant-trading-engineer + security-fortress + performance-analyzer
documentation	docs-generator + api-designer + codebase-explorer
deep-analysis	system-architect + codebase-explorer + performance-analyzer + security-fortress
database-design	database-specialist + system-architect
api-integration	api-integration-specialist + security-reviewer
realtime-feature	realtime-specialist + performance-analyzer
staff-engineer	Full lifecycle with research, planning, implementation, validation

Repository Structure

agent/
├── README.md
├── AGENT_TEMPLATE.md              # Standard template for all agents
└── prompts/
    ├── staff-engineer.md          # Core workflow agent (v2)
    ├── agent-orchestrator.md      # Multi-agent coordinator
    ├── handoff-protocol.md        # Agent communication
    ├── quick-reference.md         # Cheat sheet
    │
    ├── deep-research.md
    ├── system-architect.md
    ├── codebase-explorer.md
    ├── plan-visualizer.md
    │
    ├── security-fortress.md
    ├── security-reviewer.md
    ├── change-validator-linter.md
    ├── change-verifier.md
    ├── performance-analyzer.md
    ├── dependency-auditor.md
    │
    ├── test-generator.md
    ├── docs-generator.md
    ├── api-designer.md
    ├── refactor-assistant.md
    │
    ├── quant-trading-engineer.md
    │
    ├── domain/                    # Domain specialists
    │   ├── electron-specialist.md
    │   ├── llm-specialist.md
    │   ├── rag-specialist.md
    │   ├── database-specialist.md
    │   ├── frontend-specialist.md
    │   ├── devops-specialist.md
    │   ├── api-integration-specialist.md
    │   ├── realtime-specialist.md
    │   └── search-specialist.md
    │
    └── skills/                    # Quick commands
        ├── README.md
        ├── commit.md
        ├── review-pr.md
        ├── create-pr.md
        ├── test.md
        ├── lint.md
        ├── explain.md
        ├── debug.md
        └── refactor.md

Philosophy

• Concrete over vague: Checklists, not prose
• Safe by default: Always know how to rollback
• Right-sized process: Simple tasks get simple workflows
• Incremental: Commit working increments, validate early
• Selective: Only run relevant validators
• Domain expertise: Specialized knowledge for complex areas

Usage

These prompts can be used as:

1. System prompts when building with Claude API
2. Custom agent definitions in Claude Code (if supported)
3. Reference for guiding Claude through engineering workflows
4. Templates for building your own agents

Maintenance

Run a full local audit against installed Codex skills and this prompt repo:

./scripts/audit_codex_setup.py --skills-root ~/.codex/skills --agent-root .

Contributing

To add or improve agents:

1. Follow AGENT_TEMPLATE.md for consistent structure
2. Include Role/Trigger/Receives/Hands off header
3. Define Input/Process/Output/Handoff sections
4. Provide concrete output templates with JSON formats
5. Include quality checklists
6. Add to this README