Models, Goals and Folders
Level 2 · Inside a conversation
Change how Claude works on a task with /model, /advisor and /goal. Name a fallback model at launch. Reach more folders with /add-dir and /cd.
Course contentsLesson 6 of 17
Level 1 · Get started
Level 2 · Inside a conversation
Level 3 · Parallel work and review
Level 4 · Setup and extensions
Level 5 · Scripts and servers
After this lesson you can
- Switch model for one session only
- Set a goal Claude keeps working toward
- Choose between /add-dir and /cd
11 mincovers 6 glossary entries
Worth reading first: Context, Recap and Copy
How to change the model, set a finish line and reach more folders
/model switches the model Claude Code uses, and it saves the choice as your default for new sessions. Give a name such as opus or sonnet, or run it alone for a picker. In the picker, press s on a row to switch for this session only. The left and right arrows set the effort level for models that support it.
If the model you picked is busy or gone, --fallback-model names the next ones to try. Give a list with commas, such as sonnet,haiku, and they are tried in that order. The fallbackModel setting keeps a list for every session. The flag beats it for one session only.
/advisor turns on a second model, often a stronger one, that Claude can ask for advice. It asks at key points. That is before it picks a plan, when an error keeps coming back, and before it says the work is done. The advisor reads the whole chat and sends back advice. It is still being tested and uses more of your plan. /advisor off stops it.
/goal sets a finish line, such as all tests in test/auth pass. Claude then keeps working, turn after turn, until it is met. After each turn, a separate small model reads the chat and judges the goal. It reads only what Claude has shown, so write a goal that Claude's own output can prove.
A goal ends in four ways. It is met, the checker judges it cannot be met, an error needs you, or you run /goal clear. A session has one goal at a time. To cap it, put a limit in the goal itself, such as or stop after 20 turns.
/add-dir gives the session one more folder to read and edit, and keeps the first. Most .claude settings in the added folder do not load, so its own skills and rules stay out. --add-dir does the same at launch. /cd instead moves the whole session to a new folder and keeps the chat.
The choice is about where the work lives. A shared library next to your app wants /add-dir, because the app is still the main job. A move from the web folder to the api folder for the rest of the day wants /cd. Both offer folder names as you type, and Tab picks one.
/model opus
/advisor sonnet
/goal all tests in test/auth pass and lint is clean
/add-dir ../shared-lib
/cd ../api
# at launch, in the shell
claude --fallback-model sonnet,haikuWorked example
Lucy leaves a goal running over lunch
Lucy's branch has 9 failing tests. She types /goal all tests in test/auth pass and goes to lunch.
Claude works through 14 turns, and the checker reads the test output after each one. On turn 14 it reads 0 failures and the goal ends. Lucy comes back to a green run and a chat that shows every step.
The words for this lesson6
Check yourself
Lucy wants a second folder in this session, and to keep working in her first one. Which command fits?
Finished this lesson?
Marking it complete also marks its 6 glossary entries learned, which updates your roadmap and your saved page. 0 are already marked.