Imported from llcoolblaze/claude-boris (docs/AGENTS.md). Install upstream with npx skills add llcoolblaze/claude-boris --skill docs. Copyright stays with the author.

Agents Reference

All specialist agents in Claude Boris.

The Orchestrator

Boris

File: .claude/agents/boris.md

The master orchestrator that coordinates everything. Invoked via /boris command.

Specialties:

• Planning and task breakdown
• Delegating to specialists
• Coordinating verification
• Managing workflows

When to use: Any complex, multi-step task.

Specialist Agents

Code Architect

File: .claude/agents/code-architect.md

Senior architect for design and system decisions.

Specialties:

• Design reviews
• Architecture decisions
• System design
• Technical planning

When to use:

• Before major implementations
• When evaluating approaches
• For architectural questions
• During technical planning

Code Simplifier

File: .claude/agents/code-simplifier.md

Cleans up and simplifies code after implementation.

Specialties:

• Reducing complexity
• Improving readability
• Removing duplication
• Modernizing patterns

When to use:

• After completing a feature
• Before code review
• When code feels messy
• During refactoring

Critical rule: Never changes functionality—only improves code quality.

Test Writer

File: .claude/agents/test-writer.md

Generates comprehensive tests.

Specialties:

• Unit tests
• Integration tests
• Component tests
• Test patterns (AAA)

When to use:

• New features need tests
• Coverage is low
• Critical paths need testing
• TDD workflow

Verify App

File: .claude/agents/verify-app.md

End-to-end application verification.

Specialties:

• Build verification
• Test suite execution
• Security checks
• Environment validation

When to use:

• Before shipping features
• Before merging PRs
• After major changes
• Pre-deployment checks

PR Reviewer

File: .claude/agents/pr-reviewer.md

Automated code review.

Specialties:

• Security review
• Correctness checks
• Quality assessment
• Improvement suggestions

When to use:

• Before merging any PR
• Self-review before requesting human review
• Automated review in CI

Doc Generator

File: .claude/agents/doc-generator.md

Creates and updates documentation.

Specialties:

• README updates
• API documentation
• Component docs
• CLAUDE.md updates

When to use:

• After significant changes
• When adding new features
• When docs are outdated
• To document learnings

Oncall Guide

File: .claude/agents/oncall-guide.md

Debugs production issues.

Specialties:

• Incident response
• Rapid diagnosis
• Quick fixes
• Post-mortems

When to use:

• Something is broken in production
• Investigating errors
• Performance issues
• System outages

How Agents Work

Invoking Agents

Boris automatically invokes agents via the Task tool:

Boris: I'll have code-architect review the design first.
[Invokes code-architect agent]

Architect: Here's my analysis...

Boris: Now let's implement. [Does work]

Boris: Let me have code-simplifier clean this up.
[Invokes code-simplifier agent]

Direct Invocation

You can also ask Claude to use an agent directly:

User: Use the code-architect agent to review my API design

User: Have the test-writer create tests for the auth module

User: Run the verify-app agent before I merge

Creating Custom Agents

Add files to .claude/agents/:

---
name: my-agent
description: What this agent specializes in
tools: Read, Edit, Bash, etc.
---

# Agent Name

You are a [role] who specializes in [domain].

## Responsibilities
- Thing 1
- Thing 2

## Process
1. Step 1
2. Step 2

## Output Format
How results should look

Tips:

• Give clear role identity
• Define specific process
• Specify output format
• List what tools are needed