SMARTER USAGE · SMALL HABITS, BETTER WORK
Your ambition.
A clearer plan.
Your ambition has no limits. Your plan does. Give AI a clear target, the right context and a way to prove it worked.
BEFORE YOUR NEXT TASK
Know what’s left.
Check the remaining weekly capacity, any five-hour window and the next reset in INMC. If a reading is unavailable, refresh or reconnect before relying on it. Plan the next task around the window you actually have.
It is remaining plan capacity from the latest reading, not money or an API credit balance. Weekly and five-hour windows can reset at different times. Missing data is never treated as 100% available.
Planning in Claude Code · Open the official guide ↗
01
Give the task a finish line.
Name the outcome, the relevant input and what must stay unchanged. Start with one deliverable you can check.
Explore this habit →TRY THIS · CODEX OR CLAUDE CODE
Fix the checkout bug
Investigate why the checkout button does not respond. Read the checkout component and its event handler first. Keep the payment flow and styles unchanged. Propose the smallest fix and one check that reproduces the problem and verifies the result. Ask for missing evidence before making changes.
Replace the inputs with your own. Paste into your AI tool and review the scope before running.02
Bring the context that matters.
Keep related work together. For a different objective, carry over a short handoff instead of the whole conversation.
Explore this habit →TRY THIS · CODEX OR CLAUDE CODE
Start the next task with a clean brief
Prepare a handoff for a new conversation: objective, relevant files or sources, decisions already made, verified results and remaining questions. Keep it under 200 words. Distinguish facts from assumptions. Do not include credentials or unrelated history.
Replace the inputs with your own. Paste into your AI tool and review the scope before running.03
Ask for proof. Then move on.
Set a check before adding more work. Use the reasoning level offered by your tool that fits the task; deeper reasoning and extra agents can use more capacity.
Explore this habit →TRY THIS · CODEX OR CLAUDE CODE
Review a result without expanding scope
Review this result against the original objective. Identify what is verified, what is still uncertain and the smallest next check needed. Do not expand the scope or run unrelated tests. If the agreed checks pass, summarize the result and stop. Objective: [paste]. Result and evidence: [paste].
Replace the inputs with your own. Paste into your AI tool and review the scope before running.The official reference.
Examples above are written by INMC. Check each tool’s documentation for current behavior and plan requirements.
Claude Code
Best practices ↗Usage and costs ↗A little context.
Do these habits guarantee lower usage?
No. Usage depends on the model, reasoning, context, tools and task. These habits make work easier to scope and verify; they do not guarantee savings.
What does the INMC percentage mean?
It is remaining plan capacity from the latest reading, not money or an API credit balance. Weekly and five-hour windows can reset at different times. Missing data is never treated as 100% available.