Advanced⚙️ Code Required🧪 Experimental

Archetypes

Work mode bundles that compose CLAUDE.md, Skills, and Hooks into coherent operating modes for different types of work.

🧪

This Is Still Evolving

I'm still figuring out the best patterns for work mode archetypes. The framework below is my current working theory, but it changes as I use it. What I'm sharing is a snapshot of an ongoing experiment, not a finished system.

If you're exploring similar patterns, I'd love to hear what's working for you.

What Are Archetypes?

Archetypes are work mode bundles that compose everything you've learned so far: CLAUDE.md context, Skills workflows, and Hooks automation—into coherent operating modes.

The idea: different types of work require different mindsets. Client delivery work needs scope awareness and clear communication. Foundation work needs quality focus. Learning work needs room for experimentation.

The question archetypes answer: Given the type of work I'm doing, what context should Claude have, what skills should be active, and what guardrails should be in place?

The Four Archetypes

These map roughly to the project lifecycle: Discover → Build → Deliver → Relate

🧭

Strategist

"Ship the right thing"

Signals:

  • •Ambiguous requirements
  • •Greenfield projects
  • •"What should we build?"

Typical skills:

Requirements analysisArchitecture decisionsTrade-off evaluation
⚙️

Engineer

"Ship it right"

Signals:

  • •Code review requests
  • •Security concerns
  • •Quality focus

Typical skills:

Code reviewTesting strategyPerformance optimization
🚀

Operator

"Ship it reliably"

Signals:

  • •Deploy requests
  • •Timeline discussions
  • •Launch planning

Typical skills:

Deployment checklistScope trackingTimeline management
💼

Advisor

"Ship it valuably"

Signals:

  • •Client communications
  • •Proposals needed
  • •Difficult conversations

Typical skills:

Client communicationScope negotiationValue articulation

Project-Type Defaults

In my setup, a SessionStart hook detects the project type from the path and suggests appropriate archetypes:

Project TypeDefault ArchetypesWhy
Client workOperator + AdvisorDelivery and relationship focus
Foundation workEngineer + OperatorQuality and delivery focus
Learning workEngineerExperimentation focus

The hook suggests, but doesn't force. You can always override: "switch to Engineer", "activate Advisor", or just tell Claude what mode you need.

How It Works

Here's the flow when you open a project:

1.
SessionStart hook fires

Detects project type from path (CLIENT_WORK/, FOUNDATION/, LEARNING/)

2.
Suggests appropriate archetypes

"This looks like Operator + Advisor territory. Activate?"

3.
User confirms or adjusts

"Yes" / "Switch to Engineer" / Just start working

4.
Skills load, constraints activate

Relevant skills become active, appropriate guardrails engage

The Full Stack

Archetypes sit at the top of the Context Engineering stack:

ArchetypesWork modes

Bundle everything below into coherent operating modes

HooksAutomation

Detect context, trigger skills, enforce gates

SkillsWorkflows

Reusable prompt modules encoding expertise

CLAUDE.mdContext

Project-specific knowledge and memory

What I'm Still Figuring Out

  • →Multi-archetype blending: When work needs both Engineer and Advisor, how do skills compose?
  • →Archetype transitions: When should Claude suggest switching modes mid-session?
  • →Skill conflicts: What happens when two active skills give contradictory guidance?
  • →Right level of automation: Where's the line between helpful and over-engineered?
← HooksBack to Overview