Each profile keeps its own progress: XP, badges, certifications and reviews.
＋
New profile
Create an account
by Koneetiv · course edition · 100% local, no sign-in required · Français
🆘 Get unstuck in 30 seconds
Claude is going off track
Press Esc to stop it immediately (the session stays intact), then redirect it with a precise message. If the damage is done, /rewind restores the code and/or the conversation.
I broke my project
Don't panic, it can be undone. /rewind (or Esc twice) restores an earlier state. If you had asked Claude to make a commit: “go back to the last commit”. Good habit for next time: “make a commit” before any big change.
Claude forgot an important instruction
An instruction given in the conversation can get lost as the conversation grows. Lasting instructions belong in CLAUDE.md (open it with /memory): they are re-read at every session. For an action that must happen every single time, use a hook.
The session is getting slow or confused
The context is probably full. /compact summarises it to free up space; for a new topic, use /clear and start fresh. The rule: one conversation = one task.
My hook isn't firing
Three checks: 1) it is the right event (PreToolUse ≠ PostToolUse); 2) the matcher targets the right tool (e.g. Write); 3) the session was restarted after settings.json was changed. Ask Claude: “test my hook and tell me why it isn't firing”.
A connector has stopped working
Open /mcp to see its status. Most often the authentication has expired → sign in again. Also check that access has not been revoked on the service side (Google, Slack…).
Claude keeps asking me for permission
That is normal at first: it is the cautious mode. To speed things up, add targeted allow rules for actions you understand and consider safe: “add an allow rule for listing files”.
The response got cut off or I hit a limit
You have hit either your plan's usage limit or the context window limit. Options: wait for your usage limit to reset, use /compact, or move simple tasks to a lighter model with /model.
Which model should I choose?
Complex or critical task → Opus. Everyday work → Sonnet. Simple, repetitive, high-volume work → Haiku. Switch at any time with /model.
I'm running out of usage too fast
A long context is re-read with every message: /clear between tasks, focused conversations, and the right model for the right task. /cost shows where your usage is going.
My artifact won't run
Check that “Code execution and file creation” is turned on in your claude.ai settings (Settings → Capabilities). Then ask for it explicitly: “make it an artifact”.
I don't know how to word my request
A structure that works: the context (where, about what), the precise task, the constraints, the expected format. And the most effective shortcut: ask Claude “what information do you need to do this well?”.
🤖 Still stuck? Ask Claude itself
The best troubleshooter for Claude is Claude. Describe your problem, then copy the formatted request into your Claude Code session, or open it directly in claude.ai, already filled in.