/agent - Universal CODITECT Agent Invoker

Invoke any of the 130+ CODITECT agents by name without needing individual Python wrappers, skills, or complex dispatchers.

System Prompt

⚠️ EXECUTION DIRECTIVE: When the user invokes this command, you MUST:

1. IMMEDIATELY execute - no questions, no explanations first
2. ALWAYS show full output from script/tool execution
3. ALWAYS provide summary after execution completes

DO NOT:

• Say "I don't need to take action" - you ALWAYS execute when invoked
• Ask for confirmation unless requires_confirmation: true in frontmatter
• Skip execution even if it seems redundant - run it anyway

The user invoking the command IS the confirmation.

---

Usage

/agent <agent-name> <task-description>

Examples

# Git workflow operations
/agent git-workflow-orchestrator sync all submodules to master

# Documentation tasks
/agent codi-documentation-writer update the README.md with new component counts

# Code review
/agent code-reviewer review the changes in src/auth/ for security issues

# Multi-agent coordination
/agent orchestrator coordinate implementation of user authentication feature

# Security audit
/agent security-specialist audit the API endpoints for OWASP vulnerabilities

# DevOps tasks
/agent devops-engineer setup CI/CD pipeline for the frontend module

List Available Agents

/agent --list

Get Agent Info

/agent --info <agent-name>

How It Works

This command uses the universal invocation pattern that bridges CODITECT's 130+ custom agents with Claude Code's Task tool:

/agent <name> <task>
    ↓
invoke-agent.py reads agents/<name>.md
    ↓
Extracts system prompt and capabilities
    ↓
Task(subagent_type="general-purpose",
     prompt="You are <agent>. <system_prompt>. Task: <task>")

Key Insight: Claude Code's Task tool has ~52 built-in subagent_type values. CODITECT agents are NOT built-in types. This command uses general-purpose as a proxy, injecting the agent's system prompt into the task prompt.

Architecture

CODITECT Custom Agents (130+)          Claude Code Task Tool
├── agents/git-workflow-orchestrator.md    subagent_type:
├── agents/codi-documentation-writer.md  → - general-purpose (PROXY)
├── agents/orchestrator.md                  - devops-engineer (built-in)
├── agents/security-specialist.md           - cloud-architect (built-in)
└── ... (127 more)                          - ... (49 more built-in)

                    ↓
            invoke-agent.py
                    ↓
    Task(subagent_type="general-purpose",
         prompt="Use {agent} subagent to {task}...")

When to Use This vs Built-in Agents

Scenario	Use
CODITECT custom agent (130+)	/agent <name> <task>
Claude Code built-in agent	Task(subagent_type="<type>")
Intelligent multi-agent routing	agent_dispatcher.py
Simple single-agent task	/agent or direct Task()

Available Agent Categories

Category	Count	Examples
Development	25+	code-reviewer, debugger, frontend-development-agent
DevOps	15+	devops-engineer, cicd-automation, cloud-architect
Documentation	10+	codi-documentation-writer, documentation-generation
Security	8+	security-specialist, penetration-testing-agent
QA/Testing	8+	codi-qa-specialist, codi-test-engineer
Research	6+	market-researcher, competitive-analyst
Orchestration	5+	orchestrator, git-workflow-orchestrator

Run /agent --list for complete inventory.

Script Location

scripts/core/invoke-agent.py

Direct Script Usage

# From coditect-core root
python3 scripts/core/invoke-agent.py git-workflow-orchestrator "sync all submodules"
python3 scripts/core/invoke-agent.py --list
python3 scripts/core/invoke-agent.py --info orchestrator
python3 scripts/core/invoke-agent.py --json git-workflow-orchestrator "sync all"

Success Output

When agent invocation completes successfully:

✅ COMMAND COMPLETE: /agent <name>
Agent: <agent-name>
Task: <task-description>
Result: Task completed successfully

Completion Checklist

Before marking complete:

• Agent file found and loaded
• System prompt extracted
• Task executed by agent
• Result returned to user

Failure Indicators

This command has FAILED if:

• ❌ Agent file not found in agents/
• ❌ Agent name misspelled
• ❌ Task prompt empty
• ❌ Agent execution timed out

When NOT to Use

Do NOT use when:

• Task matches Claude Code built-in agent (use Task() directly)
• Need multi-agent orchestration (use /agent orchestrator)
• Agent doesn't exist (check with /agent --list first)

Anti-Patterns (Avoid)

Anti-Pattern	Problem	Solution
Wrong agent name	File not found	Use /agent --list
Empty task prompt	No action taken	Provide clear task description
Built-in vs custom confusion	Suboptimal routing	Check if built-in exists first

Principles

This command embodies:

• #1 Recycle, Extend, Re-Use - Leverages 130+ existing agents
• #4 Separation of Concerns - Each agent has focused capability
• #5 No Assumptions - Verify agent exists before invoking

Full Standard: CODITECT-STANDARD-AUTOMATION.md

---

Version: 1.0.0 Created: 2025-12-22 Author: CODITECT Team