Chapter C.2 — Appendices
Closing — reference material. Product details verified on 24/06/2026 against official sources.
How to use these appendices
They aren’t read in sequence: they’re consulted. You’ll find a glossary of the recurring terms, a troubleshooting table for the most common problems, and the official links from which to verify the volatile data.
Glossary
- Agentic: a way of working in which Claude carries a task forward over several steps, not just a single answer.
- API: programmatic access to the models, to integrate Claude into your own software (ch. L6.6).
- Canvas: Claude Design’s canvas where the generated interfaces appear (ch. L4.1).
- Connector: a link that gives Claude access to an app, with your permissions (ch. L3.3).
- Context window: the working memory of a single chat, measured in tokens (ch. L6.4).
- Cowork: non-code agentic work in the desktop app, in an isolated VM (ch. L3.1).
- Design /
/design-sync: the canvas and the command that syncs design and code both ways (ch. L4.3). - Dispatch: starting from your phone a task that Claude runs on your computer (ch. L6.3).
- Effort level: how deeply Claude reasons; higher consumes more usage (ch. L6.4).
- Handoff: passing a design to Claude Code so it becomes software (ch. L4.3).
- Hook: a command attached to a Claude Code event (ch. L6.1).
- MCP: open standard that connects Claude to external tools and data (ch. L6.2).
- Native installer: the recommended method to install Claude Code, without Node (ch. L2.2).
- OAuth: delegated access via the browser, without typing the password in the terminal (ch. L2.3).
- Prompt: the request you write to Claude (ch. L1.2).
- Project: a workspace with instructions and a knowledge base shared across chats (ch. L3.2).
- Routine: a Claude Code automation that runs on the cloud (ch. L6.3).
- Scheduled Task: a recurring Cowork task, runs with the computer on (ch. L6.3).
- Skill /
SKILL.md: reusable knowledge that Claude loads on its own (ch. L5.1). - Sub-agent: an assistant with its own context and permissions, that Code delegates to (ch. L6.1).
- Tool use: the use, via API, of tools you provide (ch. L6.6).
- Usage limit: how much you can use Claude in a period (ch. L6.4).
- Isolated VM: the virtual machine Cowork runs in; it sees only the connected folders (ch. L3.1).
Troubleshooting
Table C.2.1 — Common problems and first solution.
| Problem | Cause | Solution |
|---|---|---|
command not found |
PATH | new terminal (L2.2) |
| Login failed | extra API key | unset the key (L2.3) |
| Function missing | admin | Admin > Capabilities (L1.3) |
| Hook doesn’t block | exit code | use exit 2 (L6.1) |
| Connector inert | not activated | toggle in chat (L3.3) |
| Task skipped | PC off | reopen the app (L6.3) |
| Chat saturated | length limit | new chat (L6.4) |
Official links
From here you verify the data that changes over time. These are the only sources this book relies on for product details.
- claude.ai — access to Claude (web).
- claude.com and claude.com/pricing — products and plans.
- support.claude.com — official guides and troubleshooting.
- code.claude.com/docs — Claude Code documentation.
- platform.claude.com/docs — API and Console documentation.
- agentskills.io — the open standard for Skills.
- github.com/anthropics/skills — example skills.
Legal notes
Independent publication, not affiliated with or endorsed by Anthropic. Claude and Anthropic are trademarks of their respective owners. The product details (versions, commands, prices, menus, plans) are verified as of the date indicated in each chapter and in the ledger, and are subject to change: for current values refer to the official sources above and to the companion repo.
Summary
- The glossary collects the recurring terms, with the cross-reference to the chapter.
- The troubleshooting table gives the first move for common problems.
- The official links are the only sources for the volatile data.
- The publication is independent and not affiliated with Anthropic.
- The data is verified as of the date indicated and should be rechecked against the sources.
The end
You’ve gone through the whole journey: from your first messages in chat all the way to the API and automation. The product will keep changing — that’s why the volatile data is dated and collected in the companion repo. The underlying ideas, on the other hand, stay: describe the result, give context, iterate, and make the pieces of the ecosystem work together.
Reference material. The glossary and cross-references are evergreen; troubleshooting, links, and product details verified on 24/06/2026 against the official sources listed.