Claude Cert Snacks

Snack 01 – Agentic Architecture Basics (≈500 Wörter)

Snack 02 – Task Decomposition: wann splitten, wann linear lassen (≈500 Wörter)

Snack 03 – Planning-Pattern: ReAct, Plan-then-Execute, Hybrid

Snack 04 – Memory-Typen: Short-Term, Long-Term, External

Snack 05 – Context-Budgeting: Halluzinationen vs. Overload

Snack 06 – Single-Agent vs. Multi-Agent: Kosten, Latenz, Failure Surface

Snack 07 – Weekly Review + Exam-Fragetypen Woche 1

Snack 08 – Tool-Design-Prinzipien (idempotent, typed inputs, clear errors)

Snack 09 – MCP-Grundlagen: wann sinnvoll?

Snack 10 – Tool-Sicherheit: Least Privilege, Allowlists, Guardrails

Snack 11: Claude Code Setup – Repo Hygiene, Prompts, Reproducibility

Snack 12 – Workflow-Patterns: triage → execute → validate → summarize

Snack 13 – Failure Handling: Retries, Fallback, Human Handoff

Snack 14 – Weekly Review + Szenariofragen Woche 2

Snack 15 – Prompt-Struktur für komplexe Aufgaben

Snack 16 – Constraints, Roles & Rubric-based Prompting

Snack 17 – JSON/Schema-first Denken

Snack 18 – Output Validation + Auto-Repair Loops

Snack 19 – Evaluation-Prompts: Quality Checks als eigener Schritt

Snack 20 – Anti-Patterns: Overprompting, Ambiguity, Hidden Assumptions

Snack 21 – Weekly Review + Szenariofragen (Woche 3: Prompt Engineering & Structured Outputs)

Snack 22 – Reliability in Production: Timeouts, Retries, Circuit Breaker

Snack 23 – Observability: Logs, Traces, Error Taxonomy

Snack 24: Security & Injection Defense in agentischen Systemen

Snack 25 – Cost Engineering: Token-Budget, Model Tiering, Guardrails