oh-my-claude-code

Imported from zephyrpersonal/oh-my-claude-code (AGENTS.md). Install upstream with npx skills add zephyrpersonal/oh-my-claude-code. Copyright stays with the author.

Agent Documentation

Complete guide to all agents in oh-my-claude-code.

Table of Contents

• Orchestrator
• Explore
• Librarian
• Oracle
• Frontend UI/UX Engineer
• Document Writer
• Multimodal Looker

Orchestrator

File: .claude/agents/orchestrator.md Model: inherit (uses main conversation model) Cost: EXPENSIVE Category: advisor

Purpose

The orchestrator is the main coordination agent. It classifies tasks and delegates to appropriate specialists.

When to Use

• Complex multi-step tasks
• Tasks requiring coordination across multiple domains
• When you're unsure which agent to use
• Full-feature implementation with research

Key Triggers

Complex multi-step tasks → Use orchestrator

What It Does

1. Intent Classification - Analyzes request type
2. Task Planning - Creates detailed todo lists
3. Agent Selection - Routes to appropriate specialist
4. Result Verification - Checks delegated work

Example Usage

# Complex feature requiring research
claude "Use the orchestrator to add JWT authentication"

# Multi-domain task
claude "Use the orchestrator to refactor the data layer and update UI"

Explore

File: .claude/agents/explore.md Model: anthropic/claude-haiku-4-5 Cost: FREE Category: exploration

Purpose

Fast, focused codebase search. Answers "Where is X?", "Which file has Y?", "Find the code that does Z".

When to Use

• Multiple search angles needed
• Unfamiliar module structure
• Cross-layer pattern discovery
• "Where is X implemented?"
• "Which files contain Y?"

When NOT to Use

• You know exactly what to search
• Single keyword/pattern suffices
• Known file location

Key Triggers

2+ modules involved → fire explore

Output Format

<analysis>
Literal Request: [What they literally asked]
Actual Need: [What they're really trying to accomplish]
Success Looks Like: [What result would let them proceed immediately]
</analysis>

<results>
<files>
- /absolute/path/to/file1.ts — [why this file is relevant]
- /absolute/path/to/file2.ts — [why this file is relevant]
</files>

<answer>
[Direct answer to their actual need]
</answer>

<next_steps>
[What they should do with this information]
</next_steps>
</results>

Thoroughness Levels

Example Usage

claude "Use the explore agent to find all error handling patterns (medium thoroughness)"

Librarian

File: .claude/agents/librarian.md Model: anthropic/claude-sonnet-4-5 Cost: CHEAP Category: exploration

Purpose

External documentation and open-source research. Finds evidence with permalinks.

When to Use

• "How do I use [library]?"
• "What's the best practice for [framework feature]?"
• "Find examples of [library] usage"
• Weird/unexpected behavior from a library
• Need official documentation
• Need OSS implementation examples

When NOT to Use

• Internal codebase questions (use explore instead)
• Simple syntax questions
• Questions about well-known standard libraries

Key Triggers

External library mentioned → fire librarian

Output Format

Every claim includes evidence with permalinks:

**Claim**: [What you're asserting]

**Evidence** ([source name](permalink)):
\`\`\`[language]
[source code snippet]
\`\`\`

**Context**: [Why this matters]

Example Usage

claude "Use the librarian agent to find React hooks best practices with useEffect"

Oracle

File: .claude/agents/oracle.md Model: anthropic/claude-opus-4-5 Cost: EXPENSIVE Category: advisor

Purpose

Deep reasoning for architecture decisions, code analysis, and engineering guidance.

When to Use

• Complex architecture design
• After completing significant work (self-review)
• 2+ failed fix attempts
• Unfamiliar code patterns
• Security/performance concerns
• Multi-system tradeoffs

When NOT to Use

• Simple file operations
• First attempt at any fix
• Questions answerable from code you've read
• Trivial decisions (variable names, formatting)
• Things you can infer from existing code patterns

Key Triggers

2+ failed fixes or complex architecture → Consult Oracle

Output Format

## Bottom Line
[2-3 sentences capturing your recommendation]

## Action Plan
1. [First step]
2. [Second step]
...

## Effort Estimate
Quick | Short | Medium | Large

Effort Scale

Example Usage

# Architecture decision
claude "Use the oracle agent to review this database schema for scalability"

# After failed attempts
claude "I've tried 3 times to fix this memory leak. Use the oracle agent to help."

Frontend UI/UX Engineer

File: .claude/agents/frontend-ui-ux-engineer.md Model: anthropic/claude-sonnet-4-5 Cost: CHEAP Category: specialist

Purpose

Visual/UI specialist for styling, layout, animation, and responsive design.

When to Use

• Styling changes (colors, spacing, typography)
• Layout and positioning
• Animations and transitions
• Responsive design issues
• Hover states and interactions
• Visual polish and refinement

When NOT to Use

• Pure logic changes (API calls, data fetching, state management)
• Type definitions
• Utility functions
• Business logic

Key Triggers

Frontend VISUAL changes → delegate to frontend-ui-ux-engineer

Visual Keywords

If ANY of these keywords are involved, this agent handles the work:

style, className, css, tailwind, color, background, border,
shadow, margin, padding, width, height, flex, grid, animation,
transition, hover, responsive, font, text, spacing, layout

Scope

Example Usage

claude "Use the frontend-ui-ux-engineer agent to improve this button's hover state"
claude "Use the frontend-ui-ux-engineer agent to make this layout responsive"

Document Writer

File: .claude/agents/document-writer.md Model: anthropic/claude-sonnet-4-5 Cost: CHEAP Category: specialist

Purpose

Technical writing specialist for clear, accurate documentation.

When to Use

• Creating or updating README
• Writing API documentation
• Creating architecture documentation
• Writing user guides
• Adding code comments
• Creating changelogs
• Writing technical specifications

When NOT to Use

• Code implementation (main agent task)
• Simple inline comments (implementer should add)
• Non-technical writing

Key Triggers

Documentation tasks → delegate to document-writer

Documentation Types

Example Usage

claude "Use the document-writer agent to create API documentation for this module"
claude "Use the document-writer agent to update the README with installation instructions"

Multimodal Looker

File: .claude/agents/multimodal-looker.md Model: anthropic/claude-sonnet-4-5 Cost: CHEAP Category: specialist

Purpose

Visual analysis specialist for PDFs, images, screenshots, and diagrams.

When to Use

• Analyzing PDF documents
• Interpreting screenshots
• Understanding diagrams and flowcharts
• Reading charts and graphs
• Extracting text from images
• Analyzing UI mockups
• Error screenshot analysis

When NOT to Use

• Code-only tasks (use explore)
• Text-based documentation (use librarian or read directly)
• Questions that don't involve visual media

Key Triggers

PDF/image/screenshot → delegate to multimodal-looker

Media Types Handled

Example Usage

claude "Use the multimodal-looker agent to analyze this error screenshot"
claude "Use the multimodal-looker agent to extract data from this chart"

Cost-Based Decision Framework

Direct Tools (FREE) → explore (FREE) → librarian (CHEAP) → specialist → oracle (EXPENSIVE)

1. Direct Tools First: If you know exactly what to do
2. Explore Agent: Internal codebase search
3. Librarian Agent: External docs and research
4. Specialist Agents: Domain-specific work
5. Oracle Agent: Complex architecture and deep analysis

Agent Selection Quick Reference