Brite Claude Plugins

A Process + Org plugin for Claude Code. Superpowers methodology + compound engineering + Linear integration — structured workflow (brainstorm → plan → worktree → execute → review → compound → audit) with Linear woven into every step.

Versions: per-plugin — see marketplace.json | Changelog | Roadmap

Philosophy

This plugin teaches how to work, not what to know about specific technologies:

session-start → brainstorm → plan → [worktree] → execute (subagent + TDD) → review → ship (compound + audit)
       ↑                                                                                         |
       └───────────────────── scope (creative discovery) ←───────────────────────────────────────┘

• Process: Superpowers' full workflow with TDD, subagent-per-task execution, and compound knowledge
• Org: Linear integration at every step, security hooks, team conventions
• Domain skills are separate plugins — tech-stack knowledge lives in dedicated domain plugins

Influenced by superpowers and compound-engineering. See the Roadmap for what's coming next.

Vision: Brite Agent Platform

This plugin is evolving into a four-layer platform for universal project routing, domain-specific knowledge, structured workflows, and autonomous execution.

See the Platform Design Document for the full PRD.

Prerequisites

• Claude Code CLI installed
• Node.js 18+ (for MCP servers)

Quick Start

Register the marketplace, then install the plugin you want:

claude plugin marketplace add Brite-Nites/brite-claude-plugins
claude plugin install workflows@brite-claude-plugins

The other plugins in this repo (marketing, revops, cadence, flow-architecture, brite-core) install the same way — swap the name before @. See CONTRIBUTING.md for the full install and configuration reference.

Verify installation by typing /workflows: in Claude Code — you should see the available commands in the slash menu.

Plugins in this repo

This is a multi-plugin bundle. workflows is the primary Process + Org plugin (its commands are listed below); the others are domain plugins that install the same way. Versions are per-plugin — see marketplace.json.

Available Commands

The commands below belong to the workflows plugin.

Core workflow (the inner loop):

Direction-setting:

Utilities:

Quality:

Compound knowledge & governance:

Skill Coverage Matrix

Skills activate automatically when Claude detects relevant context.

Inner Loop skills (auto-activate in sequence):

Design, backend & quality skills (shipped):

Review Agents

/workflows:review runs up to 9 specialized review agents in parallel:

Depth modes: fast (Tier 1 only), thorough (default, Tier 1+2), comprehensive (all tiers). All agents run on Opus. A Haiku-powered diff-triage agent gates trivial diffs.

Workflow skills (shipped):

MCP Servers

The workflows plugin configures three MCP servers automatically:

The Linear MCP server provides tools for managing issues, projects, milestones, and documentation directly from Claude Code.

gbrain-team needs GBRAIN_CLIENT_ID and GBRAIN_CLIENT_SECRET exported in your shell profile before you launch Claude Code, and GBRAIN_WRITE_CLIENT_ID / GBRAIN_WRITE_CLIENT_SECRET as well if you have write access. Setup steps are in CONTRIBUTING.md § Team gbrain credentials; the reasoning is ADR-045.

Plugin Development Guide

Repository Structure

.claude-plugin/
  marketplace.json          # Plugin registry (required for bundles)
plugins/
  workflows/
    .claude-plugin/
      plugin.json           # Plugin metadata
    commands/               # Slash commands
    skills/                 # Model-invoked skills
    agents/                 # Specialized subagents
    hooks/                  # Event handlers
    .mcp.json               # MCP server configurations

Plugin Manifest (plugin.json)

Each plugin requires .claude-plugin/plugin.json:

{
  "name": "workflows",
  "description": "Process + Org plugin — structured workflow methodology with Linear integration",
  "version": "3.24.0",
  "author": { "name": "Brite" },
  "homepage": "https://github.com/brite-nites/brite-claude-plugins",
  "repository": "https://github.com/brite-nites/brite-claude-plugins",
  "license": "MIT",
  "keywords": ["claude-code", "plugin", "process", "workflow", "linear"],
  "commands": "./commands/",
  "skills": "./skills/"
}

Adding Commands

Commands are slash commands that users invoke directly. Create a markdown file in commands/.

File Structure

commands/
  my-command.md      # Becomes /workflows:my-command

Command Format

---
description: Brief description shown in slash menu
---

Instructions for Claude on how to handle this command.

Use $ARGUMENTS for the full user input after the command.
Use $1, $2, etc. for positional arguments.

Example: Deploy Command

commands/deploy.md:

---
description: Deploy the current project to production
---

You are deploying the user's project. Follow these steps:

1. Run all tests to ensure the build is stable
2. Build the production bundle
3. Deploy using the configured deployment method

The user specified: $ARGUMENTS

Adding Skills

Skills are model-invoked capabilities that Claude uses automatically based on context. Unlike commands, users don't invoke skills directly — Claude decides when to use them.

File Structure

skills/
  code-review/
    SKILL.md           # Required: Skill definition
    templates/         # Optional: Supporting files
  security-scan/
    SKILL.md

SKILL.md Format

---
name: code-review
description: When to use this skill - Claude reads this to decide
allowed-tools: Read, Grep, Glob
user-invocable: true
---

# Code Review Skill

Detailed instructions for how Claude should perform this skill.

## When to Activate
- User asks for code review
- PR review is requested
- Code quality assessment needed

## Process
1. First step...
2. Second step...

Frontmatter Options

Adding MCP Server Configurations

MCP (Model Context Protocol) servers connect Claude to external tools, APIs, and databases.

File Location

Create .mcp.json at the plugin root:

plugins/
  workflows/
    .mcp.json           # MCP configuration

Configuration Format

{
  "mcpServers": {
    "database": {
      "command": "npx",
      "args": ["@brite/mcp-database"],
      "env": {
        "DB_PATH": "${CLAUDE_PLUGIN_ROOT}/data"
      }
    },
    "api-client": {
      "command": "${CLAUDE_PLUGIN_ROOT}/servers/api-server.js",
      "args": ["--config", "${CLAUDE_PLUGIN_ROOT}/config.json"]
    }
  }
}

Path Variables

Transport Types

stdio (default):

{
  "mcpServers": {
    "local-server": {
      "command": "./server.js",
      "args": []
    }
  }
}

HTTP/SSE:

{
  "mcpServers": {
    "remote-server": {
      "url": "https://api.example.com/mcp",
      "transport": "sse"
    }
  }
}

Adding Hooks

Hooks respond to Claude Code events automatically (formatting, validation, notifications).

File Location

Create hooks/hooks.json or define inline in plugin.json:

plugins/
  workflows/
    hooks/
      hooks.json

Hook Events

Configuration Format

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Write|Edit",
        "hooks": [
          {
            "type": "command",
            "command": "${CLAUDE_PLUGIN_ROOT}/scripts/format.sh $CLAUDE_FILE_PATH"
          }
        ]
      }
    ],
    "PreToolUse": [
      {
        "matcher": "Bash",
        "hooks": [
          {
            "type": "prompt",
            "prompt": "Verify this command is safe to run"
          }
        ]
      }
    ]
  }
}

Hook Types

Adding Agents

Agents are specialized subagents that Claude can delegate tasks to.

File Structure

agents/
  security-reviewer.md
  performance-tester.md

Agent Format

---
description: What this agent specializes in
capabilities:
  - security analysis
  - vulnerability detection
  - compliance checking
skills: security-scan, code-review
---

# Security Reviewer Agent

You are a security specialist. When delegated tasks:

1. Review code for security vulnerabilities
2. Check for OWASP Top 10 issues
3. Validate input sanitization
4. Report findings with severity levels

Adding Output Styles

Output styles customize Claude's response format and behavior.

File Structure

styles/
  concise.md
  verbose.md

Style Format

---
name: Concise Mode
description: Short, direct responses
keep-coding-instructions: true
---

Respond concisely. Use bullet points. Avoid lengthy explanations.
Maximum 3 sentences per response unless more detail is explicitly requested.

Marketplace Configuration

The root marketplace.json registers plugins for distribution:

{
  "name": "brite-claude-plugins",
  "owner": {
    "name": "Brite"
  },
  "metadata": {
    "description": "Claude Code plugins for the Brite organization"
  },
  "plugins": [
    {
      "name": "workflows",
      "source": "./plugins/workflows",
      "description": "Process + Org plugin — structured workflow methodology with Linear integration",
      "version": "3.24.0"
    }
  ]
}

To add a new plugin to the bundle, create the plugin directory and add it to the plugins array.

Documentation

Contributing

See CONTRIBUTING.md for the full contributor guide, including:

• Step-by-step instructions for adding commands, skills, agents, and hooks
• SKILL.md frontmatter standard
• Branch naming and commit message conventions
• CI checks and local testing

License

MIT

Issue Tracking

Issues for this project are tracked in Brite Skill Packs (Layer C). Sibling Linear projects under the same initiative: Brite Orchestration Layer, Brite Knowledge Layer, Brite Runtime & Harness.