Models are pattern engines
A language model predicts useful responses from patterns in its training and the context you provide. It can reason impressively, but it can also produce confident mistakes. Give it evidence and review the result.
See events
AI foundations
Learn enough to build something useful.
Use this as a practical map—not homework. Start with the six ideas below, then open the official references when you want to go deeper.
Choose a workshop track ↗Context is the working brief
Your website, business facts, tone, rules, examples, task input, and conversation history make up the working context. Better context usually matters more than clever wording.
A prompt is an instruction
A useful prompt makes the task, inputs, expected result, boundaries, and review criteria clear. Start with a real task, test the output, and improve the weak spots you can observe.
A Skill makes it reusable
A Skill packages instructions and supporting resources for a specific repeatable job. It helps Claude apply the same workflow again without rebuilding your guidance every time.
Connectors provide access
Connectors let Claude reach approved apps, files, and services. A Skill teaches the procedure; a connector supplies data or actions. Always review permissions and keep human approval around consequential actions.
Agents choose their next steps
An agent can plan, use tools, inspect results, and continue across multiple steps. That flexibility adds cost and risk, so begin with a narrow workflow and add autonomy only when it clearly helps.
Strong fit
Use AI to draft, structure, compare, transform, and explore.
AI is especially useful when a human can describe the task, provide the source material, recognize a good result, and review the output before it matters.
Use caution
Do not outsource truth, judgment, permission, or accountability.
Keep humans responsible for sensitive decisions, commitments, legal or medical advice, money movement, confidential data, and changes to real systems.
Official references
Keep learning after the workshop.
These links go to Anthropic’s current documentation and guidance. Product interfaces can change, so use the linked instructions as the source of truth.
Start here
Prompt engineering overview
Define success, test a first draft, and improve what the evidence says is weak.
Open reference ↗ Working memoryHow context windows work
Understand what Claude can reference in a conversation and why focused context matters.
Open reference ↗ Workshop outcomeWhat are Skills?
A plain-language explanation of reusable Skills and how they differ from projects and connectors.
Open reference ↗ How-toCreate a custom Skill
Anthropic’s official instructions for structuring, packaging, and testing a Skill.
Open reference ↗ After the workshopUse Skills in Claude
Enable, upload, inspect, invoke, and troubleshoot the Skills in your account.
Open reference ↗ ExtensionsUse connectors with Claude
See how Claude can reach apps and services, and why permissions deserve careful review.
Open reference ↗ Deeper diveBuilding effective agents
Learn the difference between fixed workflows and agents that dynamically choose their steps.
Open reference ↗ PrivacySensitive data in Claude
Review Anthropic’s current consumer guidance before using sensitive information in chats.
Open reference ↗