Workflow Plugin: Claude Workflow V2
A universal Claude Code workflow plugin with 7 agents, 26 commands, 14 skills, and 14 hooks for any software project.
Evidence shows: README describes permission templates (settings.local.json.template) and MCP templates, but does not explicitly mention least privilege; hooks include security scan and file protection, but user confirmation mechanism is not described; data flow transparency is insufficient, telemetry collection not explained; sensitive data handling only mentions blocking secrets, not encryption or storage; external effects (e.g., auto-push, PR creation) lack explicit user confirmation; rollback only mentioned for dependency-upgrade command, but mechanism not detailed; source attribution only mentions creator, not verified. Deductions: most security mechanisms are only described, lacking implementation details and verification.
Evidence shows: README structure is clear, command/agent/skill/hook lists are consistent, but no actual runtime verification; dependency availability does not mention dependency locking or version pinning; failure messages lack specific error handling examples. Deductions: static review cannot verify runtime behavior, and documentation lacks failure handling details.
Evidence shows: README describes multiple usage scenarios (commands, agents, skills, hooks) and provides installation options and compatibility notes; capability boundaries are described via commands and agents, but not explicitly limited; trigger precision is described via auto-triggers, but precise rules not provided; environment fit mentions Claude Code, Cursor, etc., but detailed configuration not provided. Deductions: some descriptions are generic, lacking specific configuration examples.
Evidence shows: README provides clear directory structure and installation instructions; naming is stable, command and agent names are consistent; examples and FAQ section provides usage examples, but FAQ is missing; known limitations not mentioned; license is MIT, but no version changelog; maintenance responsibility not clear. Deductions: missing known limitations, changelog, and maintenance responsibility.
Evidence shows: README provides usage examples for commands, agents, skills, hooks, with clear output formats; marginal value is high, offering multiple automation features; cost-benefit not explicit, but installation and usage cost is low. Deductions: cost-benefit not quantified, and static review cannot verify actual effectiveness.
Evidence shows: README claims (e.g., command count, agent count) are consistent with directory structure, but no independent verification; cross-source verification insufficient, relying only on README; facts and inferences not clearly separated. Deductions: lack of independent verification and fact/inference separation.
- Static review cannot verify actual runtime behavior; all scores are based on documentation and code structure.
- Security mechanisms (e.g., permissions, user confirmation, data flow) are only described, lacking implementation details and verification.
- Missing known limitations, changelog, and maintenance responsibility, which may affect long-term use.
What does this agent do, and when should you use it?
claude-workflow-v2 is a comprehensive Claude Code plugin that bundles 7 specialized agents (e.g., code reviewer, debugger, security auditor), 26 slash commands covering git workflows, verification, parallel review, and planning, 14 on-demand skills (e.g., architecture design, performance optimization), and 14 automation hooks (security blocking, auto-formatting, desktop notifications). It integrates with 35+ AI coding tools via skills.sh, but the full plugin (agents, commands, hooks) is exclusive to Claude Code. The plugin is MIT-licensed and can be installed via npx or by cloning the repository.
The plugin extends Claude Code's capabilities: after installation, /project-starter:commit auto-generates conventional commit messages, /project-starter:commit-push-pr performs a full commit-push-PR workflow, and /project-starter:verify-changes spawns multiple subagents to validate build, tests, linting, and security. When a user request matches a trigger (e.g., "the login is broken"), the debugger agent activates and traces the issue. Hooks run automatically on events, such as blocking edits that contain secrets or formatting files with prettier/black. Skills load domain knowledge on demand, such as optimizing performance or designing APIs. All interactions are achieved through reading files, executing shell commands, and producing structured outputs.
- A developer wants to automate routine git tasks, such as generating commit messages, pushing, and creating pull requests.
- A team wants to run multi-agent verification on every change, including build, tests, lint, and security scans.
- A user needs to prevent accidental commits of secrets by having edits blocked automatically when API keys are detected.
- A developer wants to load domain knowledge (e.g., architecture patterns, performance optimization) on demand within Claude Code sessions.
- A team wants desktop notifications for task completion or when Claude needs input, streamlining the workflow.
- A user wants to reuse the plugin's skills inside other tools like Cursor or Codex.
What are this agent's strengths and limitations?
- Comprehensive component set: 7 agents, 26 commands, 14 skills, 14 hooks covering the entire development lifecycle.
- High automation: hooks automatically run security scans and formatting on events like edits and session start.
- Multi-agent orchestration: supports parallel review and verification to catch issues early.
- Skills are portable across 38+ AI coding tools via skills.sh.
- MIT licensed, free to use and modify.
- Full functionality (agents, commands, hooks) is limited to Claude Code; other tools only get skills.
- Requires Claude Code v1.0.33+, so not compatible with older versions.
- Hooks and scripts depend on Python 3 and Git, adding environment requirements.
- Installation and configuration involve copying multiple template files, which can be complex for beginners.
- Some commands (e.g.,
commit-push-pr) require GitHub CLI or Git credentials, adding setup overhead.
How do you install or deploy this agent?
Recommended: run npx skills add CloudAI-X/claude-workflow-v2 to install skills to Claude Code, Cursor, Codex, and 35+ other agents. For the full plugin (agents, commands, hooks) exclusive to Claude Code, run npx install-claude-workflow-v2@latest. Alternatively, clone the repo and run claude --plugin-dir ./claude-workflow-v2 per session, or use claude plugin install for permanent installation. Requires Claude Code v1.0.33+.
How do you use this agent?
After installation, type /project-starter:architect to enter architecture-first mode, /project-starter:commit to auto-commit changes, or /project-starter:verify-changes to run multi-agent verification. Customize by adding Markdown files in commands/, agents/, and skills/ directories. Optionally, copy templates/settings.local.json.template to your project's .claude directory to pre-approve common commands, copy CLAUDE.md.template for team conventions, and configure MCP servers for integrations like Slack or GitHub.
How does this agent compare with similar options?
The main alternative is Claude Code's built-in capabilities or other plugins. This plugin differentiates by the sheer number of components and automation, but you should evaluate whether its workflow fits your team's existing practices.
FAQ
Does this plugin cost money?
What permissions does the plugin request?
What if installation fails?
/plugin to see if the plugin is listed and check the Errors tab. Ensure Claude Code is v1.0.33+ and that Python 3, Git, and Node.js (optional) are installed.Does it support tools other than Claude Code?
How do I add custom commands or agents?
commands/ or agents/ directories following the documented format. Skills follow the SKILL.md convention in the skills/ directory.