Files
Claude-Code-Workflow/docs/agents/builtin.md
catlog22 c3ddf7e322 docs: add VitePress documentation site
- Add docs directory with VitePress configuration
- Add GitHub Actions workflow for docs build and deploy
- Support bilingual (English/Chinese) documentation
- Include search, custom theme, and responsive design
2026-02-28 16:14:09 +08:00

14 KiB

Built-in Agents

CCW includes 21 specialized agents organized across 5 categories, each designed for specific development tasks. Agents can work independently or be orchestrated together for complex workflows.

Categories Overview

Category Count Primary Use
CLI 6 CLI-based interactions, exploration, and planning
Development 5 Code implementation and debugging
Planning 4 Strategic planning and issue management
Testing 3 Test generation, execution, and quality assurance
Documentation 3 Documentation and design systems

CLI Agents

cli-explore-agent

Purpose: Specialized CLI exploration with 3 analysis modes

Capabilities:

  • Quick-scan (Bash only)
  • Deep-scan (Bash + Gemini)
  • Dependency-map (graph construction)
  • 4-phase workflow: Task Understanding → Analysis Execution → Schema Validation → Output Generation

Tools: Bash, Read, Grep, Glob, ccw cli (gemini/qwen/codex), ACE search_context

Task({
  subagent_type: "cli-explore-agent",
  prompt: "Analyze authentication module dependencies"
})

cli-discuss-agent

Purpose: Multi-CLI collaborative discussion with cross-verification

Capabilities:

  • 5-phase workflow: Context Prep → CLI Execution → Cross-Verify → Synthesize → Output
  • Loads discussion history
  • Maintains context across sessions

Tools: Read, Grep, Glob, ccw cli

Calls: cli-explore-agent for codebase discovery before discussions

Task({
  subagent_type: "cli-discuss-agent",
  prompt: "Discuss architecture patterns for microservices"
})

cli-execution-agent

Purpose: Intelligent CLI execution with automated context discovery

Capabilities:

  • 5-phase workflow: Task Understanding → Context Discovery → Prompt Enhancement → Tool Execution → Output Routing
  • Background execution support
  • Result polling

Tools: Bash, Read, Grep, Glob, ccw cli, TaskOutput

Calls: cli-explore-agent for discovery before execution

Task({
  subagent_type: "cli-execution-agent",
  prompt: "Execute security scan on authentication module"
})

cli-lite-planning-agent

Purpose: Lightweight planning for quick task breakdowns

Capabilities:

  • Creates simplified task JSONs without complex schema validation
  • For straightforward implementation tasks

Tools: Read, Write, Bash, Grep

Task({
  subagent_type: "cli-lite-planning-agent",
  prompt: "Plan user registration feature"
})

cli-planning-agent

Purpose: Full-featured planning for complex implementations

Capabilities:

  • 6-field schema with context loading
  • Flow control and artifact integration
  • Comprehensive task JSON generation

Tools: Read, Write, Bash, Grep, Glob, mcp__ace-tool__search_context

Task({
  subagent_type: "cli-planning-agent",
  prompt: "Plan microservices architecture migration"
})

cli-roadmap-plan-agent

Purpose: Strategic planning for roadmap and milestone generation

Capabilities:

  • Creates long-term project plans
  • Generates epics, milestones, and delivery timelines
  • Issue creation via ccw

Tools: Read, Write, Bash, Grep

Task({
  subagent_type: "cli-roadmap-plan-agent",
  prompt: "Create Q1 roadmap for payment system"
})

Development Agents

code-developer

Purpose: Core code execution for any implementation task

Capabilities:

  • Adapts to any domain while maintaining quality standards
  • Supports analysis, implementation, documentation, research
  • Complex multi-step workflows

Tools: Read, Edit, Write, Bash, Grep, Glob, Task, mcp__ccw-tools__edit_file, mcp__ccw-tools__write_file

Task({
  subagent_type: "code-developer",
  prompt: "Implement user authentication with JWT"
})

tdd-developer

Purpose: TDD-aware code execution for Red-Green-Refactor workflows

Capabilities:

  • Extends code-developer with TDD cycle awareness
  • Automatic test-fix iteration
  • CLI session resumption

Tools: Read, Edit, Write, Bash, Grep, Glob, ccw cli

Extends: code-developer

Task({
  subagent_type: "tdd-developer",
  prompt: "Implement payment processing with TDD"
})

context-search-agent

Purpose: Specialized context collector for brainstorming workflows

Capabilities:

  • Analyzes existing codebase
  • Identifies patterns
  • Generates standardized context packages

Tools: mcp__ace-tool__search_context, mcp__ccw-tools__smart_search, Read, Grep, Glob, Bash

Task({
  subagent_type: "context-search-agent",
  prompt: "Gather context for API refactoring"
})

debug-explore-agent

Purpose: Debugging specialist for code analysis and problem diagnosis

Capabilities:

  • Hypothesis-driven debugging with NDJSON logging
  • CLI-assisted analysis
  • Iterative verification
  • Traces execution flow, identifies failure points, analyzes state at failure

Tools: Read, Grep, Bash, ccw cli

Workflow: Bug Analysis → Hypothesis Generation → Instrumentation → Log Analysis → Fix Verification

Task({
  subagent_type: "debug-explore-agent",
  prompt: "Debug memory leak in connection handler"
})

universal-executor

Purpose: Versatile execution for implementing any task efficiently

Capabilities:

  • Adapts to any domain while maintaining quality standards
  • Handles analysis, implementation, documentation, research
  • Complex multi-step workflows

Tools: Read, Edit, Write, Bash, Grep, Glob, Task, mcp__ace-tool__search_context, mcp__exa__web_search_exa

Task({
  subagent_type: "universal-executor",
  prompt: "Implement GraphQL API with authentication"
})

Planning Agents

action-planning-agent

Purpose: Pure execution agent for creating implementation plans

Capabilities:

  • Transforms requirements and brainstorming artifacts into structured plans
  • Quantified deliverables and measurable acceptance criteria
  • Control flags for execution modes

Tools: Read, Write, Bash, Grep, Glob, mcp__ace-tool__search_context, mcp__ccw-tools__smart_search

Task({
  subagent_type: "action-planning-agent",
  prompt: "Create implementation plan for user dashboard"
})

conceptual-planning-agent

Purpose: High-level planning for architectural and conceptual design

Capabilities:

  • Creates system designs
  • Architecture patterns
  • Technical strategies

Tools: Read, Write, Bash, Grep, ccw cli

Task({
  subagent_type: "conceptual-planning-agent",
  prompt: "Design event-driven architecture for order system"
})

issue-plan-agent

Purpose: Issue resolution planning with closed-loop exploration

Capabilities:

  • Analyzes issues and generates solution plans
  • Creates task JSONs with dependencies and acceptance criteria
  • 5-phase tasks from exploration to solution

Tools: Read, Write, Bash, Grep, mcp__ace-tool__search_context

Task({
  subagent_type: "issue-plan-agent",
  prompt: "Plan resolution for issue #123"
})

issue-queue-agent

Purpose: Solution ordering agent for queue formation

Capabilities:

  • Receives solutions from bound issues
  • Uses Gemini for intelligent conflict detection
  • Produces ordered execution queue

Tools: Read, Write, Bash, ccw cli (gemini), mcp__ace-tool__search_context, mcp__ccw-tools__smart_search

Calls: issue-plan-agent

Task({
  subagent_type: "issue-queue-agent",
  prompt: "Form execution queue for issues #101, #102, #103"
})

Testing Agents

test-action-planning-agent

Purpose: Specialized agent for test planning documents

Capabilities:

  • Extends action-planning-agent for test planning
  • Progressive L0-L3 test layers (Static, Unit, Integration, E2E)
  • AI code issue detection (L0.5) with CRITICAL/ERROR/WARNING severity
  • Project-specific templates
  • Test anti-pattern detection with quality gates

Tools: Read, Write, Bash, Grep, Glob

Extends: action-planning-agent

Task({
  subagent_type: "test-action-planning-agent",
  prompt: "Create test plan for payment module"
})

test-context-search-agent

Purpose: Specialized context collector for test generation workflows

Capabilities:

  • Analyzes test coverage
  • Identifies missing tests
  • Loads implementation context from source sessions
  • Generates standardized test-context packages

Tools: mcp__ccw-tools__codex_lens, Read, Glob, Bash, Grep

Task({
  subagent_type: "test-context-search-agent",
  prompt: "Gather test context for authentication module"
})

test-fix-agent

Purpose: Execute tests, diagnose failures, and fix code until all tests pass

Capabilities:

  • Multi-layered test execution (L0-L3)
  • Analyzes failures and modifies source code
  • Quality gate for passing tests

Tools: Bash, Read, Edit, Write, Grep, ccw cli

Task({
  subagent_type: "test-fix-agent",
  prompt: "Run tests for user service and fix failures"
})

Documentation Agents

doc-generator

Purpose: Documentation generation for technical docs, API references, and code comments

Capabilities:

  • Synthesizes context from multiple sources
  • Produces comprehensive documentation
  • Flow_control-based task execution

Tools: Read, Write, Bash, Grep, Glob

Task({
  subagent_type: "doc-generator",
  prompt: "Generate API documentation for REST endpoints"
})

memory-bridge

Purpose: Documentation update coordinator for complex projects

Capabilities:

  • Orchestrates parallel CLAUDE.md updates
  • Uses ccw tool exec update_module_claude
  • Processes every module path

Tools: Bash, ccw tool exec, TodoWrite

Task({
  subagent_type: "memory-bridge",
  prompt: "Update CLAUDE.md for all modules"
})

ui-design-agent

Purpose: UI design token management and prototype generation

Capabilities:

  • W3C Design Tokens Format compliance
  • State-based component definitions (default, hover, focus, active, disabled)
  • Complete component library coverage (12+ interactive components)
  • Animation-component state integration
  • WCAG AA compliance validation
  • Token-driven prototype generation

Tools: Read, Write, Edit, Bash, mcp__exa__web_search_exa, mcp__exa__get_code_context_exa

Task({
  subagent_type: "ui-design-agent",
  prompt: "Generate design tokens for dashboard components"
})

Orchestration Patterns

Agents can be combined using these orchestration patterns:

Inheritance Chain

Agent extends another agent's capabilities:

Parent Child Extension
code-developer tdd-developer Adds TDD Red-Green-Refactor workflow, test-fix cycle
action-planning-agent test-action-planning-agent Adds L0-L3 test layers, AI issue detection

Sequential Delegation

Agent calls another agent for preprocessing:

Caller Callee Purpose
cli-discuss-agent cli-explore-agent Codebase discovery before discussion
cli-execution-agent cli-explore-agent Discovery before CLI command execution

Queue Formation

Agent collects outputs from multiple agents and orders them:

Collector Source Purpose
issue-queue-agent issue-plan-agent Collect solutions, detect conflicts, produce ordered queue

Context Loading Chain

Agent generates context packages used by execution agents:

Context Provider Consumer Purpose
context-search-agent code-developer Provides brainstorming context packages
test-context-search-agent test-fix-agent Provides test context packages

Quality Gate Chain

Sequential execution through validation gates:

code-developer (IMPL-001)
  → test-fix-agent (IMPL-001.3 validation)
  → test-fix-agent (IMPL-001.5 review)
  → test-fix-agent (IMPL-002 fix)

Agent Selection Guide

Task Recommended Agent Alternative
Explore codebase cli-explore-agent context-search-agent
Implement code code-developer tdd-developer
Debug issues debug-explore-agent cli-execution-agent
Plan implementation cli-planning-agent action-planning-agent
Generate tests test-action-planning-agent test-fix-agent
Review code test-fix-agent doc-generator
Create documentation doc-generator ui-design-agent
UI design ui-design-agent -
Manage issues issue-plan-agent issue-queue-agent

Tool Dependencies

Core Tools

All agents have access to: Read, Write, Edit, Bash, Grep, Glob

MCP Tools

Specialized agents use: mcp__ace-tool__search_context, mcp__ccw-tools__smart_search, mcp__ccw-tools__edit_file, mcp__ccw-tools__write_file, mcp__ccw-tools__codex_lens, mcp__exa__web_search_exa

CLI Tools

CLI-capable agents use: ccw cli, ccw tool exec

Workflow Tools

Coordinating agents use: Task, TaskCreate, TaskUpdate, TaskList, TaskOutput, TodoWrite, SendMessage

::: info See Also