Loop

Prompting Foundations

The techniques that reliably steer Claude, from clarity to agentic behavior.

0 of 7 lessons complete · 91 min total

Start module
  1. Be Clear, Direct, and SpecificTreating Claude like a brilliant new hire with no context: explicit success criteria, audience, and format beat vague requests every time.Lesson · 14 minNot completed
  2. Multishot: Teaching by Example3-5 diverse, well-chosen examples outperform paragraphs of instructions — and how to spot when examples encode the wrong pattern.Lesson · 13 minNot completed
  3. Letting Claude ThinkChain of thought and adaptive thinking: when reasoning room improves accuracy, and when it just adds latency.Lesson · 14 minNot completed
  4. XML Tags, Roles & Long-Context StructureStructuring prompts with tags, system-role framing, and placing long documents before instructions for reliable long-context performance.Lesson · 14 minNot completed
  5. Prefilling Is Dead; Chaining Lives OnWhy response prefilling no longer works with modern models and how chained subtask prompts replace it cleanly.Lesson · 13 minNot completed
  6. Steering Agentic BehaviorTuning action defaults, parallelism, and over-engineering tendencies when Claude acts rather than answers.Lesson · 14 minNot completed
  7. Prompting CheckpointTechnique selection, example design, and structure for long contexts.Checkpoint · 9 minNot completed