Claude & AI Workflows
Patterns for working effectively with AI coding agents — from context engineering to workflow architectures.
This section documents patterns, research findings, and practical frameworks for working effectively with AI coding agents. The goal is to move beyond ad-hoc prompting toward systematic, repeatable workflows that compound in value over time.
Start Here
Claude Code Setup
CLAUDE.md, rules, hooks, and session management
Skills Catalogue
Browse all available automation skills
Context Engineering
The discipline that replaced prompt engineering
Concepts
Context Engineering
Strategies for curating optimal tokens during LLM inference
CLAUDE.md Playbook
The primary interface between your codebase and an AI agent
Agent Efficiency Research
What works and what doesn't — reconciling the studies
Research to Validation
Five-step process keeping agent work traceable
Practical
Claude Code Setup
Configure Claude Code on a new project
Skill-Based Automation
Patterns for creating reusable skills
Skills Catalogue
Directory of all available skills
Workflow Architectures
Command, Agent, Skill patterns and beyond
Guiding Principles
- Minimal over maximal. Less context, delivered precisely, outperforms walls of text.
- Human-curated over generated. Agent-written documentation about how to use agents creates a feedback loop that degrades quality.
- Compounding over static. The best context files grow through real corrections, not speculative additions.
- Traceable over ad-hoc. Every prototype should connect back to the research that motivated it.