CloudYetiFREE
Claude Code

Claude Code tips from the team that builds Claude Code

The habits and commands from the videos, each with the text to copy. A new tip is added with each new video.

1. Claude Code now reads AGENTS.md: keep one instruction file for every agent

Since version 2.1.277, Claude Code reads AGENTS.md when a project has no CLAUDE.md. If you also use Codex or another agent, one AGENTS.md file can hold the rules for all of them.

claude -v

Full steps for this tip →

2. Put /goal inside /loop, and Claude Code works through your task file

One command makes Claude Code wake up on a timer, read tasks.md, and keep working until each pending task is done. You add tasks to the file instead of typing prompts.

/loop 2m /goal check tasks.md and implement every pending task

Full steps for this tip →

3. An old Claude Code chat can eat a big part of your limit with one message

The prompt cache lasts one hour. After that, your next message sends the whole chat again at full price. Two commands tell you when to start fresh.

/usage
/clear

Full steps for this tip →

4. When Claude goes sideways, the Claude Code team stops and re-plans

When a Claude Code task starts going wrong, switch back to plan mode and fix the plan instead of pushing more prompts. Then have a second Claude review the plan.

Review this plan as a staff engineer. List what could break, what is missing, and which steps need a test before we build.

Full steps for this tip →

5. The Claude Code team keeps context usage on screen, all the time

One /statusline command puts a bar at the bottom of Claude Code that shows how full the context window is. A second one adds your 5-hour and weekly plan limits.

/statusline show model and context percentage with a progress bar
/statusline also show my 5-hour and weekly limit used

Full steps for this tip →

6. The two words the Claude Code team adds to big requests

End a big Claude Code request with "use subagents". Claude splits the work into helpers that each run in their own context window, so your main chat stays clean.

use subagents

Full steps for this tip →

7. The line the Claude Code team types after every correction

When Claude Code gets something wrong and you fix it, one line makes Claude write a rule so the same mistake does not come back.

Update your CLAUDE.md so you don't make that mistake again.

Full steps for this tip →

8. 7 learning swaps for AI engineers who use coding agents

Keep learning to code and debug. Then move your study time from syntax drills and tutorial copying to the skills that decide whether a coding agent's work is right.

1. Less syntax drills, more system design: where data lives and how failures spread.
2. Less prompt tricks, more context engineering: the right files, tools and constraints.
3. Less vague requests, more clear specifications: the result and the tests before the agent starts.
4. Less tutorial copying, more agent environments: tools, permissions and isolated workspaces.
5. Less looks-good reviews, more verification: run the app, test failures, check what changed.
6. Less watching every step, more reliable task loops: checkpoints, retry limits and a human to escalate to.
7. Less leaderboard chasing, more cost and product judgment: cost per successful task and value to users.

Full steps for this tip →

Free live lesson · Nov 24Stop Measuring AI by Tokens. Measure What It EarnsOne hour, live, with Saurav. Bring your own bill. Free.
Save my free seat →
Learn it live, then run it yourselfLoop and harness engineering, AI agents on AWS, AI cost and ROI.See the workshops →
Get new videos and tips by emailOne short email when there is something new. Unsubscribe any time.

More free AI resources