ark-analysis

name	Ark Analysis
description	Analyze the Ark codebase by cloning the repository to a temporary location. Use this skill when the user asks questions about how Ark works, wants to understand Ark's implementation, or needs to examine Ark source code.

Ark Analysis

This skill helps you analyze the Ark codebase by cloning the repository and examining its contents.

When to use this skill

Use this skill when:

• User asks "how does X work in Ark?"
• User wants to understand Ark's architecture or implementation
• User needs to examine Ark source code, CRDs, or controllers
• User mentions analyzing the Ark repository

Quick start

Clone the Ark repository to a temporary location:

git clone [e-mail]:mckinsey/agents-at-scale-ark.git /tmp/ark-analysis
cd /tmp/ark-analysis

Codebase structure

The Ark repository is organized as follows:

• ark/ - Kubernetes operator and default executor (Go)

  • Controller dispatches queries to executors via A2A protocol
  • CRDs: Agent, Model, Query, Team, MCPServer, ExecutionEngine, A2AServer
  • executors/completions/ - Built-in default execution engine

• lib/ark-sdk/ - Python SDK (generated + overlay)

  • BaseExecutor ABC and ExecutorApp for pluggable executor interface

• services/ - Component services

  • ark-api/ - REST API gateway (Python/FastAPI)
  • ark-broker/ - In-memory event bus (Node.js/Express)
  • ark-dashboard/ - Web UI (Next.js/React)

• samples/ - Example YAML configurations

• docs/ - Documentation site (Next.js/MDX)

Common analysis tasks

Find controllers

ls ark/internal/controller/
grep -r "Reconcile" ark/internal/controller/

Find CRDs

ls ark/config/crd/bases/
grep -r "kind: Agent" samples/

Find A2A implementations

find . -path "*/a2a*" -type f
grep -r "A2AServer" .

Search for specific features

# Use ripgrep or grep to search
rg "query controller" --type go
grep -r "team coordination" --include="*.go"

Best practices

1. Clone to /tmp: Always clone to /tmp/ark-analysis to avoid cluttering the workspace
2. Navigate first: cd /tmp/ark-analysis before running analysis commands
3. Use search tools: Prefer rg (ripgrep) or grep for code searches
4. Check CLAUDE.md: Look for project-specific guidance in CLAUDE.md files
5. Clean up: Optionally remove the temp directory when done: rm -rf /tmp/ark-analysis

Example workflows

Analyzing a controller

git clone [e-mail]:mckinsey/agents-at-scale-ark.git /tmp/ark-analysis
cd /tmp/ark-analysis
cat ark/internal/controller/query_controller.go
grep -r "ExecuteQuery" ark/internal/genai/

Understanding A2A integration

cd /tmp/ark-analysis
find samples/a2a -name "*.py"
cat samples/a2a/simple-agent/src/simple_a2a_server/main.py
cat docs/content/developer-guide/building-a2a-servers.mdx

Finding CRD specifications

cd /tmp/ark-analysis
ls ark/api/v1alpha1/
cat ark/api/v1alpha1/agent_types.go