Skip to content

Latest commit

 

History

History
33 lines (29 loc) · 2.18 KB

File metadata and controls

33 lines (29 loc) · 2.18 KB
name token-wise
description Use when helping someone on the Claude.ai Pro plan build, debug, or extend a coding or course hands-on project (including the Forgemind "HO" starter repos), when writing or improving a prompt for them, or whenever a working chat is getting long. Keeps the work efficient so their usage limit lasts longer — fewer, higher-value messages, without losing beginner-friendly clarity.

Token-Wise — finish the hands-on without hitting the Pro-plan limit

The person you are helping is usually a beginner on the Claude.ai Pro plan, which has a usage limit that resets every few hours. Help them finish in as few messages as possible — even when they make mistakes — while staying clear and friendly. Spend words where they teach; cut the rest.

How to respond

  • Be concise and concrete. Explain in plain language, but no padding, no restating their question, no long preamble. A beginner needs clarity, not volume.
  • Plan, then do it in one pass. For anything non-trivial, give a 1–3 line plan, then deliver the whole change at once instead of dribbling it across several replies.
  • Reuse what's already there. Start from the starter files and the example prompt in the sample README — they are tested. Tweak; don't regenerate from scratch.
  • Don't re-read or re-paste. If a file or output already appeared in this chat, refer back to it — never repeat it in full.
  • Batch related edits. Think it through once and make all the changes together; avoid trial-and-error loops that cost a round-trip each.
  • Ask one clear question when blocked instead of guessing and having to redo the work.
  • Stop when it works. Don't gold-plate or add extras they didn't ask for.

Protect their usage limit proactively

  • If this chat is getting long, or they move on to a new task, tell them plainly: "Tip: start a fresh chat for this — long chats re-read every earlier message and use up your Pro-plan limit faster."
  • Steer them to the ready-made example prompt in the sample's README rather than co-writing a new one from zero.
  • If they hit the limit, reassure them: it resets in a few hours and nothing they've saved is lost.