Claude Code Skills, Agents & Workflows

Built by TRAN Le Khanh ([email protected]) — developer and AI tooling author.

Modular packages that extend Claude Code with specialized knowledge, workflows, and tool integrations.

Quick Navigation: Quick Reference | Common Workflows | Skill Index

Quick Start

CLI Installation (Recommended)

# Install globally (npm registry)
npm install -g @khanhcan148/mk

# Install kit to current project (./.claude/)
mk init

# Install kit globally (~/.claude/ — available in all projects)
mk init --global

# Preview what would be installed without writing files
mk init --dry-run

Note: This package has limited access. Email [email protected] for an invitation to access the npm package and repository.

CLI Commands

mk init              Install kit files to current project (./.claude/)
mk init --global     Install kit globally (~/.claude/)
mk init --dry-run    Preview what would be installed
mk update            Update kit files (preserves your edits by default)
mk update --force    Update and overwrite user-modified files
mk remove            Remove all kit files tracked by manifest
mk --version         Show installed CLI version
mk --help            Show help

Manual Installation (Alternative)

For contributors with repository access:

# Clone the repository (requires access to private repo)
git clone https://github.com/khanhtran148/mk-kit.git
cd mk-kit

Project-level usage (skills available in one project):

cp -r .claude /path/to/your/project/

Global-level usage (skills available across all projects):

cp -r .claude ~/.claude/

Usage

# Use a workflow command
/mk-init           # Bootstrap new project
/mk-brainstorm     # Explore options, debate trade-offs; pass Jira/AzDO URL to fetch ticket context first
/mk-plan           # Create implementation plan
/mk-implement      # End-to-end feature delivery
/mk-test           # Run tests and validate
/mk-review         # Code quality review
/mk-debug          # Debug issues
/mk-security       # Security scan and audit
/mk-db             # Database operations
/mk-docs           # Generate documentation
/mk-overview       # Multi-tier stakeholder overview (Executive, Product, Technical)
/mk-git            # Git operations
/mk-audit          # Code archaeology chain: orientation, testing, heatmap, breaking-change analysis, technical debt, domain extraction
/mk-selftest       # Self-validation checks on kit agents, skills, docs, workflows
/mk-log-analysis   # Datadog + Azure App Insights log analysis with severity triage and Mermaid visual reports
/mk-wiki           # Wiki management: staleness detection, batch refresh, bootstrap, stats, structural validation

Structure

├── .claude/
│   ├── agents/      # 36 agents (5 primary + 31 utility: implementers, quality, docs, specialized, concerns, brainstorm critics)
│   ├── skills/      # 67 skill packages (SKILL.md + scripts/references/assets)
│   │   ├── mk-*/    # 20 workflow commands (/mk-audit, /mk-brainstorm, /mk-log-analysis, /mk-overview, /mk-wiki, etc.)
│   │   └── ...      # Domain skills (frontend, backend, testing, browser automation, etc.)
│   └── workflows/   # Development protocols
├── bin/             # CLI entry point (mk command)
├── src/             # CLI source code
├── docs/            # Project documentation
│   └── plans/       # Implementation plans (created during development)
├── package.json     # npm package metadata
├── CLAUDE.md        # Claude Code guidance
└── README.md        # This file

Primary Workflow

Orchestration is handled by /mk-* skills which spawn utility agents as needed. See Common Workflows for practical examples.

Architecture:

User → /mk-* command (skill) → spawns utility agents → agents use knowledge skills

Sequential chaining:

/mk-plan → /mk-implement → /mk-test → /mk-review

Entry point commands:

| Command | Purpose | |---------|---------| | /mk-init | Full project bootstrap from conception to deployment; includes brownfield detection (Phase 0 gate) and adoption workflow (B1-B4.5 stages); generates AGENTS.md template (or migrates existing tool configs); suggests /mk-audit as next step for brownfield projects | | /mk-brainstorm | Ideation, requirements exploration, structured debate (analyzer opus agent); pass a Jira or AzDO URL to fetch ticket hierarchy first (Phase 0.5) | | /mk-audit | Code archaeology chain: orientation report, characterization testing, topology heatmap, breaking-change analysis, technical debt dashboard, domain extraction; enriches CLAUDE.md and AGENTS.md with auto-generated architecture and debt sections | | /mk-plan | Create implementation plans with phases and tasks (opus) | | /mk-design | UI/UX wireframes, design systems, mockups | | /mk-implement | End-to-end feature delivery (sonnet) | | /mk-test | Run tests, analyze coverage, validate builds | | /mk-review | Code review for quality, security, performance | | /mk-debug | Debugging workflow: investigate, diagnose, fix, verify; always offers incident journal documentation with severity hints | | /mk-security | Security-first workflow: dependency scan, secrets detection | | /mk-db | Database operations: diagnose, optimize, design, migrate | | /mk-docs | Generate and update project documentation; maintains AGENTS.md; Impact Areas analysis produces human-readable "What changed / Who is affected / What could go wrong" narrative | | /mk-git | Git operations: branch, commit, push, PR, merge | | /mk-research | Deep multi-source research on technical topics | | /mk-spike | Investigate external service integrations: fetch API docs, evaluate options, produce -spike.md with Go/No-Go | | /mk-overview | Synthesize project artifacts into multi-tier stakeholder overview: Executive Brief, Product Report, Technical Report | | /mk-workflow | Trace REST endpoint call chains with upstream caller detection, variant branching, side effects/feature flags, Mermaid diagrams | | /mk-log-analysis | Analyze production logs from Datadog or Azure Application Insights via MCP; progressive severity triage, pattern detection, mandatory stack trace investigation, mk-debug integration | | /mk-selftest | Run self-validation checks on kit agents, skills, docs, and workflows | | /mk-wiki | Manage wiki health: staleness detection, batch refresh, bootstrap, stats, structural validation |

Primary Agents

Invoked directly via /mk-* skills:

| Agent | Purpose | Model | |-------|---------|-------| | analyzer | Deep research, decision matrix, architecture diagrams, adversarial debate (Phases 5-8 of /mk-brainstorm) | opus | | planner | Implementation planning with phases, tasks, dependencies | opus | | implementer | End-to-end feature implementation | sonnet | | database-admin | Database operations: diagnose, optimize, design, migrate | sonnet | | git-manager | Git operations: commit, push, PR, merge | sonnet |

Utility Agents

Spawned by primary agents for domain-specific work:

Implementation & Quality (9) | Agent | Purpose | |-------|---------| | backend-implementer | API and domain logic implementation | | frontend-implementer | UI components, consuming API contract | | mobile-implementer | Mobile app implementation (Flutter, React Native, Swift, Kotlin) with TFD | | tester | Test execution and validation | | debugger | Issue investigation and diagnosis | | refactorer | Refactoring with TFD-first safety workflow |

Docs & Project Management (4) | Agent | Purpose | |-------|---------| | project-manager | Progress tracking, roadmap updates | | docs-manager | Documentation maintenance | | journal-writer | Technical difficulty documentation | | wiki-maintainer | Background wiki writes to docs/llm-wiki/ (spawned by mk-* skills) |

Research & Design (3) | Agent | Purpose | |-------|---------| | researcher | Technical research | | ui-ux-designer | Frontend design | | copywriter | Marketing copy and content |

Utilities & Search (2) | Agent | Purpose | |-------|---------| | scout | Codebase file search (internal) | | scout-external | Codebase search via external tools | | mcp-manager | MCP server integrations |

Parallel Concern Review Agents (11) | Agent | Purpose | |-------|---------| | quality-reviewer | Code quality, readability, maintainability (mk-review concern) | | security-reviewer | Security vulnerabilities, threat analysis (mk-review concern) | | performance-reviewer | Performance profiling, optimization opportunities (mk-review concern) | | tfd-reviewer | Test-first development verification (mk-review concern) | | dep-scanner | Dependency vulnerability scanning (mk-security concern) | | secret-scanner | Secrets and credential detection (mk-security concern) | | code-security-reviewer | Code-level security analysis (mk-security concern) | | infra-reviewer | Infrastructure security review (mk-security concern) | | log-analyzer | Log analysis and diagnostics (mk-debug concern) | | hypothesis-generator | Root cause hypothesis generation (mk-debug concern) | | log-collector | External observability platform queries — Datadog and Azure App Insights MCP (mk-log-analysis concern) |

Full agent details →

Skills

See Skill Index for a complete categorized list of all skills.

Each skill contains:

• SKILL.md (required) - Instructions (≤250 lines)
• scripts/ (optional) - Executable code with tests
• references/ (optional) - Documentation chunks
• assets/ (optional) - Templates, images

# Create new skill
python .claude/skills/skill-creator/scripts/package_skill.py <path/to/skill-folder>

Documentation

| Document | Description | |----------|-------------| | Skill Index | Categorized list of all skills and workflows | | mk-* Workflows & Agents | How each /mk-* command works and which agents it uses | | Project Overview & PDR | Requirements, constraints, success metrics | | Project Roadmap | Phases, milestones, and risk register | | Changelog | Version history and release notes | | Code Standards | Conventions, naming, quality gates | | System Architecture | Component diagrams, data flow | | Product Domain Glossary | Core domain concepts, terminology, and ecosystem definitions | | Quick Reference | Common commands and quick lookup | | Common Workflows | Practical workflow examples |

Core Principles

• YAGNI - Don't add functionality until necessary
• KISS - Prefer simple solutions
• DRY - Extract common patterns
• SOLID - For OOP/enterprise projects

Key Constraints

• SKILL.md must be ≤250 lines
• Reference files in references/ have no line limit; loaded on-demand
• Scripts must include tests
• Cross-platform compatibility (Windows + macOS/Linux) is mandatory: Scripts use Python or Node.js only — no bash/shell scripts
• Token efficiency - minimize context usage

Cross-Platform Compatibility

This kit is designed to work on Windows, macOS, and Linux:

• Scripts: All executable scripts use Python or Node.js (no bash/shell scripts)
• Claude Code Bash tool: Provides a Unix-like shell on all platforms, including Windows — git commands and other Bash tool invocations work everywhere
• External tool installs: Reference files include install instructions for macOS (brew), Ubuntu/Debian (apt-get), and Windows (winget/choco) where applicable Supported runtime: Claude Code CLI. All skills, agents, and workflows are designed for the Claude Code runtime.