Claude Code Principles Exercises — AI skill for Claude Code

Seven Principles of Agent Problem Solving: Practical Exercises

**By Panaversity -- Master the Principles That Make AI Agents Effective**

Welcome! This package contains 14 exercises + 3 capstone projects designed to teach you the Seven Principles of General Agent Problem Solving. Each principle is a pattern that separates effective AI agent directors from those who struggle. You'll practice each principle in realistic scenarios using Claude Code and Cowork.

---

What Are the Seven Principles?

These are the operational principles that make the difference between an AI agent that reliably solves problems and one that creates new problems:

1. Bash is the Key -- Use terminal commands to verify state before acting
2. Code as Universal Interface -- Express requirements as code, not prose
3. Verification as Core Step -- Always verify output systematically
4. Small, Reversible Decomposition -- Break big changes into atomic, revertable steps
5. Persisting State in Files -- Save context so nothing is lost between sessions
6. Constraints and Safety -- Guardrails enable autonomy
7. Observability -- See what your agent is doing at every step

---

Package Structure

principles-exercises/
+-- README.md                              <- You are here
+-- EXERCISE-GUIDE.md                      <- Full pedagogical guide
+-- module-1-bash-is-the-key/
|   +-- exercise-1.1-server-detective/     (investigate a project before modifying)
|   +-- exercise-1.2-deploy-disaster/      (diagnose what went wrong without bash)
+-- module-2-code-as-interface/
|   +-- exercise-2.1-report-spec/          (turn vague request into code spec)
|   +-- exercise-2.2-lost-in-translation/  (compare prose vs code instructions)
+-- module-3-verification/
|   +-- exercise-3.1-data-audit/           (systematically verify a transformation)
|   +-- exercise-3.2-silent-corruption/    (find bugs that "look right")
+-- module-4-decomposition/
|   +-- exercise-4.1-migration-ste