cab

Add a new subagent to the current project

Add a new custom slash command to the current project

Add a new skill to the current project

Stage changes, commit with concise, descriptive message, push to remote, and create a pull request in one workflow

Pull recent project activity from git log, open PRs, issues, and optionally connected MCP sources (Slack, GitHub) into a single context summary for session bootstrapping

Start a structured task using the standard execution protocol (PLAN → REVIEW → EXECUTE → VERIFY → COMMIT)

Initialize a new CAB plugin project with full structure, git setup, and optional GitHub remote

Set up git worktrees for parallel Claude Code agent execution

Integrate CC architecture into an existing project. Analyzes the current codebase, proposes tailored extensions, and scaffolds only the CC overlay without modifying existing files.

Regenerate knowledge base INDEX files from current content

Set up global user configuration in ~/.claude/

Create a new Claude Code plugin project with full structure

Scan codebase for technical debt including code duplication, unused exports, TODO/FIXME markers, and common anti-patterns. Run at end of session or periodically.

Validate current project structure against architecture standards

Expert guidance on Claude Code architecture AND active project analysis. Use for architecture advice, project structure review, or integrating existing projects with the plugin architecture. Invoke explicitly when starting architecture work.

Main orchestrator agent for multi-agent autonomous operation. Routes tasks to domain-specialist agents, synthesizes results, manages cross-session state, and enforces the standard task execution protocol (PLAN → REVIEW → EXECUTE → VERIFY → COMMIT). Set as default agent via settings.json for fully autonomous plugin operation. Use when coordinating multi-step workflows across specialists.

Analyzes existing projects and proposes Claude Code architecture integration aligned with the CAB (cc-architecture-builder) framework. Conversational, systematic, never writes without explicit user approval. Use when integrating an existing codebase with Claude Code or reviewing/restructuring current Claude Code configuration.

End-to-end verification specialist. Validates implementations against acceptance criteria using domain-appropriate testing methods. Use after implementation work is complete to confirm correctness before committing. Use PROACTIVELY after any implementation agent finishes its work.

Analyze existing codebases and design new system architectures. Covers code structure review, full-stack application design, component decomposition, data modeling, API design, technology selection, and architecture decision records (ADRs). Use when reviewing architecture, designing systems from scratch, evaluating tech stack choices, assessing scalability/security, or decomposing monoliths. Applies systems thinking across frontend, backend, data, and infrastructure layers.

Create individual Claude Code components like skills, agents, commands, or hooks. Use when user wants to add a skill, add an agent, create a command, add a hook, or extend an existing project with new components.

Enforces the standard task execution protocol (PLAN → REVIEW → EXECUTE → VERIFY → COMMIT) for non-trivial tasks. Use when implementing features, fixing bugs, performing refactors, or any multi-step work that produces artifacts requiring verification. Automatically invoked by the orchestrator agent for delegated work.

Plan and scope projects from inception through production. Decomposes requirements into phased deliverables, generates SOW (scope) and implementation plans (execution), defines acceptance criteria, and structures sprint/phase backlogs. Use when starting new projects, scoping features, writing requirements, planning sprints, or when the orchestrator needs to decompose a complex task into executable subtasks.

Generates Claude Code configuration scaffolding based on the CAB (cc-architecture-builder) base architecture. Use when user wants to quickly create CLAUDE.md, directory structure, or component templates without extensive discovery. Fast, template-driven, user refines afterward.

Create new Claude Code project structures with proper architecture. Use when user wants to create a new project, start a new plugin, initialize a custom LLM, or scaffold a project from scratch.

Validate Claude Code project structure for compliance with architecture standards. Use when user wants to validate, check, audit, or verify their project structure, or before publishing/distributing a plugin.