Claude CTX Plugin - Documentation Index

Version 0.1.0 · Last updated 2025-10-17

Claude CTX keeps your agents, commands & workflows in perfect sync.

Activate a fully curated context stack for Claude Code: purpose-built slash commands, dependency-aware agents, behavioural modes, and a Python CLI that all speak the same language. Configure once, deploy everywhere.

View on GitHub → Install the plugin

Claude Code Native Python CLI MCP Ready Blue Ops Theme

Overview

The Claude CTX Plugin is a comprehensive context management toolkit packaged as a Claude Code plugin. It provides curated agents, slash commands, behavioral modes, rules, and workflow orchestration for systematic software development.

Quick Links

Project Structure

claude-ctx-plugin/
├── commands/           # Slash command definitions (37 commands across 11 categories)
│   ├── analyze/       # Code analysis, security scanning, troubleshooting
│   ├── deploy/        # Release preparation and deployment
│   ├── design/        # System design and workflow planning
│   ├── dev/           # Development tasks (build, test, review, git, implement)
│   ├── docs/          # Documentation generation and indexing
│   ├── orchestrate/   # Multi-agent task orchestration
│   ├── quality/       # Code quality improvement and cleanup
│   ├── reasoning/     # Dynamic reasoning depth control
│   ├── session/       # Session persistence and reflection
│   ├── test/          # Test generation
│   └── tools/         # Tool selection and optimization
│
├── agents/            # Active specialized agents (11 agents)
│   ├── cloud-architect.md
│   ├── code-reviewer.md
│   ├── debugger.md
│   ├── deployment-engineer.md
│   ├── kubernetes-architect.md
│   ├── python-pro.md
│   ├── security-auditor.md
│   ├── terraform-specialist.md
│   ├── typescript-pro.md
│   ├── dependencies.map         # Agent dependency graph
│   └── triggers.yaml            # Agent activation triggers
│
├── agents-disabled/   # Disabled agents library (65 agents available)
│
├── modes/             # Active behavioral modes
│   └── Task_Management.md
│
├── modes-inactive/    # Inactive modes (can be activated as needed)
│   ├── Brainstorming.md
│   ├── Introspection.md
│   ├── Orchestration.md
│   └── Token_Efficiency.md
│
├── rules/             # Reusable rule sets
│   ├── workflow-rules.md       # Git workflow, task patterns
│   ├── quality-rules.md        # Code organization, failure investigation
│   └── efficiency-rules.md     # Tool optimization, workspace hygiene
│
├── profiles/          # Profile templates for different workflows
│   ├── enhanced/
│   └── templates/
│
├── workflows/         # Multi-agent workflow definitions
│   ├── bug-fix.yaml
│   ├── feature-development.yaml
│   ├── performance-optimize.yaml
│   └── security-audit.yaml
│
├── scenarios/         # Scenario-based orchestration definitions
│
├── claude_ctx_py/     # Python CLI implementation
│   ├── cli.py         # CLI entry point
│   ├── core.py        # Core functionality
│   └── init_cmds.py   # Command initialization
│
├── schema/            # Validation schemas
├── scripts/           # Helper scripts
│
├── CLAUDE.md          # Framework entry point
├── FLAGS.md           # Behavioral flags and execution modes
├── RULES.md           # Core behavioral rules
├── README.md          # Project overview
└── pyproject.toml     # Python package configuration

Key Capabilities

Installation

Via Claude Code Plugin System

1. Add the marketplace that references this repository:

   # See companion project: NickCrew/claude-marketplace
   

2. Install the plugin:

   /plugin install claude-ctx@<marketplace-name>
   

3. Restart Claude Code to load commands and agents

Via Python CLI

Quick Install (Recommended):

# Install everything (package, completions, manpage)
./scripts/install.sh

# Or use make
make install

Manual Installation:

# Install package (editable installs supported)
python3 -m pip install -e ".[dev]"

# Verify installation
claude-ctx --version

# View documentation
man claude-ctx

# List available modes
claude-ctx mode list

# Generate dependency map
claude-ctx agent graph --export dependency-map.md

What Gets Installed:

• Python package (claude-ctx command)
• Shell completions (bash/zsh/fish)
• Manpage (man claude-ctx)
• Development dependencies (pytest, mypy, black)

For more detailed setup guidance (including shell completion and CLI overrides) see Getting Started.

Data Directory Overrides

The CLI resolves its workspace using the following precedence:

1. CLAUDE_CTX_HOME
2. CLAUDE_PLUGIN_ROOT (automatically set when commands run inside Claude Code)
3. ~/.claude

Examples:

# Use the plugin cache that Claude Code maintains
export CLAUDE_CTX_HOME="$HOME/.claude/plugins/cache/claude-ctx"

# Or target a local checkout of this repository
export CLAUDE_CTX_HOME="$HOME/Developer/personal/claude-ctx-plugin"

claude-ctx mode status

Once exported (for example in ~/.zshrc), both the CLI and Claude Code share a single source of truth for agents, commands, and workflows.

Core Systems

1. Command System

Location: commands/ Count: 34 slash commands across 10 categories

Commands provide curated behavioral prompts for specific development tasks. Each command includes:

• Trigger conditions for automatic activation
• Usage patterns and examples
• MCP integration specifications
• Tool coordination patterns
• Success criteria and boundaries

Categories:

• /analyze - Code analysis, security, troubleshooting (6 commands)
• /deploy - Release preparation (1 command)
• /design - System design, workflow planning (2 commands)
• /dev - Development tasks (6 commands)
• /docs - Documentation generation (2 commands)
• /orchestrate - Multi-agent orchestration (3 commands)
• /quality - Code quality (2 commands)
• /session - Session management (3 commands)
• /test - Test generation (1 command)
• /tools - Tool selection (1 command)

Reference: Command Documentation

2. Agent System

Location: agents/ and agents-disabled/ Active Agents: 11 | Available: 65

Agents are specialized personas with domain expertise and specific capabilities. The system includes:

• Dependency Management: dependencies.map tracks agent relationships
• Trigger System: triggers.yaml defines automatic activation patterns
• Activation Control: Move between agents/ and agents-disabled/ to control availability

Active Agents:

• cloud-architect - Cloud infrastructure design
• code-reviewer - Code quality analysis
• debugger - Issue diagnosis and resolution
• deployment-engineer - Release engineering
• kubernetes-architect - K8s orchestration
• python-pro - Python expertise
• security-auditor - Security assessment
• terraform-specialist - Infrastructure as code
• typescript-pro - TypeScript expertise

Reference: Agent Documentation

3. Mode System

Location: modes/ and modes-inactive/ Active Modes: 1 | Available: 4

Modes are opinionated context modules that toggle workflow defaults and behavioral patterns.

Active:

• Task_Management - Multi-step task orchestration

Available:

• Brainstorming - Collaborative discovery
• Introspection - Meta-cognitive analysis
• Orchestration - Multi-tool optimization
• Token_Efficiency - Symbol-enhanced communication

Activation: Move mode files between modes/ and modes-inactive/ or use flags

4. Flag System

Location: FLAGS.md

Behavioral flags enable specific execution modes and tool selection patterns.

Categories:

• Mode Activation: --brainstorm, --introspect, --task-manage, --orchestrate
• MCP Servers: --c7, --seq, --serena, --magic, --morph, --play
• Analysis Depth: --think, --think-hard, --ultrathink
• Execution Control: --delegate, --loop, --validate, --safe-mode
• Output Optimization: --uc, --scope, --focus

5. Rule System

Location: rules/

Reusable rule sets that define behavioral constraints and best practices.

Rule Sets:

• workflow-rules.md - Task patterns, git workflow, implementation completeness
• quality-rules.md - Code organization, naming conventions, failure investigation
• efficiency-rules.md - Tool optimization, parallel operations, workspace hygiene

Priority System:

• 🔴 CRITICAL - Security, data safety (never compromise)
• 🟡 IMPORTANT - Quality, maintainability (strong preference)
• 🟢 RECOMMENDED - Optimization, style (apply when practical)

6. Workflow System

Location: workflows/

Pre-defined multi-agent sequences for common development tasks.

Available Workflows:

• feature-development.yaml - Complete feature implementation workflow
• bug-fix.yaml - Systematic bug resolution
• security-audit.yaml - Comprehensive security assessment
• performance-optimize.yaml - Performance analysis and optimization

Reference: Workflow Documentation

7. Python CLI

Location: claude_ctx_py/ Entry Point: claude-ctx

Python CLI for managing context components outside of Claude Code.

Capabilities:

• Mode management (list, activate, deactivate)
• Agent dependency visualization
• Command execution
• Context validation

Reference: CLI Documentation

Deep Dive Documentation

Comprehensive guides to claude-ctx architecture, optimization strategies, and advanced patterns.

Framework Entry Points

CLAUDE.md

Primary framework entry point that loads:

• Core framework (FLAGS.md, PRINCIPLES.md, RULES.md)
• Workflow rules (automatic for all development tasks)
• Conditional rules (loaded as needed)
• Active behavioral modes
• MCP documentation

PRINCIPLES.md

Software engineering principles and philosophy:

• Task-First Approach: Understand → Plan → Execute → Validate
• Evidence-Based Reasoning
• Parallel Thinking
• SOLID principles
• Systems Thinking
• Decision Framework

RULES.md

Core behavioral rules with priority system:

• Scope Discipline
• Professional Honesty
• Safety Rules
• Temporal Awareness

Configuration Files

Plugin Manifest

File: .claude-plugin/plugin.json

{
  "name": "claude-ctx",
  "version": "0.1.0",
  "description": "Context orchestration plugin",
  "commands": ["./commands"]
}

Python Package

File: pyproject.toml

[project]
name = "claude-ctx-py"
version = "0.1.0"
requires-python = ">=3.9"

[project.scripts]
claude-ctx = "claude_ctx_py.cli:main"

Development Workflows

Feature Development

1. Create feature branch (git checkout -b feature/name)
2. Use /design:workflow to plan implementation
3. Use /dev:implement with appropriate persona flags
4. Use /dev:code-review for quality validation
5. Use /test:generate-tests for test coverage
6. Use /dev:git for commit with semantic message
7. Use /deploy:prepare-release for deployment preparation

Bug Fixing

1. Use /analyze:troubleshoot for diagnosis
2. Use /dev:implement to fix issue
3. Use /test:generate-tests for regression coverage
4. Use /dev:code-review for validation
5. Use /dev:git to commit fix

Code Quality

1. Use /analyze:code for quality assessment
2. Use /quality:improve for systematic improvements
3. Use /quality:cleanup for dead code removal
4. Use /dev:code-review for validation

MCP Integration

The plugin integrates with Model Context Protocol servers for enhanced capabilities:

Context7

• Purpose: Official library documentation lookup
• Flag: --c7 or --context7
• Use For: Framework patterns, API documentation, version-specific implementations

Sequential

• Purpose: Multi-step reasoning and analysis
• Flag: --seq or --sequential
• Use For: Complex debugging, system design, hypothesis testing

Serena

• Purpose: Symbol operations and session persistence
• Flag: --serena
• Use For: Large codebase navigation, project memory, semantic understanding

Best Practices

Command Usage

1. Use namespace prefixes (/dev:, /test:, etc.)
2. Specify scope with flags (--scope file|module|project)
3. Focus analysis with --focus flags
4. Enable validation for critical operations (--validate)

Agent Coordination

1. Let dependency system manage agent activation
2. Use triggers.yaml for automatic agent selection
3. Compose multiple agents for complex tasks
4. Review dependencies.map for orchestration patterns

Mode Management

1. Activate modes based on task complexity
2. Use Task_Management for multi-step operations
3. Enable Token_Efficiency for large operations
4. Combine modes for optimal behavior

Flag Optimization

1. Use --safe-mode for production operations
2. Apply --think-hard for architectural decisions
3. Enable --delegate for large-scale changes
4. Use --uc for token efficiency under pressure

Troubleshooting

Commands Not Loading

1. Verify plugin installation: /plugin list
2. Check plugin manifest: .claude-plugin/plugin.json
3. Restart Claude Code
4. Validate command syntax in markdown files

Agent Not Activating

1. Check agent location (agents/ vs agents-disabled/)
2. Review triggers.yaml for activation conditions
3. Verify dependencies in dependencies.map
4. Manually invoke with /agent activate <name>

CLI Issues

1. Verify installation: python3 -m pip show claude-ctx-py
2. Check Python version: python3 --version (>=3.9 required)
3. Reinstall: python3 -m pip install --force-reinstall .

Contributing

Adding Commands

1. Choose appropriate namespace under commands/
2. Create markdown file following template structure
3. Add to namespace README.md
4. Update command count in this index

Creating Agents

1. Create markdown file in agents-disabled/
2. Define triggers in triggers.yaml
3. Add dependencies to dependencies.map
4. Move to agents/ when ready for activation

Defining Workflows

1. Create YAML file in workflows/
2. Define trigger conditions and steps
3. Specify agent sequence and success criteria
4. Update workflows/README.md

References

Internal Documentation

• Architecture Guide - System design and patterns
• Agent Catalog - Complete agent reference with 78 agents
• Agent Skills Guide - Progressive disclosure and token optimization
• Model Optimization - Haiku vs Sonnet strategy
• Command Reference - Complete command catalog
• Getting Started - Installation and setup

External Resources

• Claude Code Documentation
• Plugin Development Guide
• MCP Specification

Version History

0.1.0 (2025-10-17)

• Initial plugin release
• 34 slash commands across 10 categories
• 11 active agents, 65 total available
• 4 behavioral modes
• 4 pre-defined workflows
• Python CLI for context management
• Complete documentation system

Support

Repository: github.com/NickCrew/claude-ctx-plugin

Issues: Report a bug or request a feature

Marketplace: claude-marketplace

This documentation was generated using /docs:index command