Effective habits

The difference between casual users and power users isn't secret prompts; it's a handful of habits applied consistently. Here they are in one place.

The core habits

Brief, don't search. Give role, task, context, format, and constraints. Treat Claude like a colleague you're delegating to.
Show an example. One sample of "good" output beats paragraphs of description.
Iterate. The first answer is a draft. Give specific feedback; converge in 2–3 rounds.
Right-size the model. Sonnet by default, Opus for hard things, Haiku for fast/simple. (Choosing.)
Let it ask you questions. "Ask me 3 clarifying questions first" prevents wasted effort.
Verify what matters. Check facts, numbers, and code before you rely on them.
Systematize repetition. Promote good one-offs into templates, Projects, and Skills. (Repeatable.)

When the answer is generic → add context. Ask: "What did I leave Claude to guess?"

When the format is wrong → specify it explicitly, or show an example.

When it's a hard problem → ask Claude to think step by step and consider alternatives before concluding.

When you're stuck on a decision → have Claude red-team your thinking.

When a task recurs → stop re-typing; build a Project or Skill.

To actually improve over time:

Keep a prompt swipe file. When something works great, save it.
Notice your repetition. Once a week, ask what you did more than once, and systematize it.
Review failures. When Claude disappoints, identify which habit you skipped.

Give Claude what a smart colleague would need, check its work, and turn what works into a system.

Master that and you've mastered Claude.

Next, two essential guardrails: privacy & data and limitations & trust.