anaksor-ocr
Sokheng · Kit
Reads Khmer and English text from documents, receipts and photos of signs, using a detector for text lines and a recogniser for each line. Suits developers who…

Rewrites a prompt to ask for the same result in fewer tokens. Every requirement, name and format stays.
Prompt diet is an AI prompt skill in the Developer category, listed by Mindket. It is free. This is version 1, last updated on . Copy its prompt from this page into any AI chat and put your text where it says {{input}}.
Rewrite this prompt so it asks for exactly the same result in as few words as possible. Keep every requirement, constraint, name, number and output format. Cut filler, repetition and politeness. Return the rewritten prompt, then one line: "About N words before, M after."
{{input}}Prompt diet rewrites a prompt so it asks for the same result in fewer words. Every requirement, constraint, name, number and output format you wrote stays. Greetings, repeats and polite padding go. On a prompt that goes out with every call, the saving repeats each time.
Anyone who pays per token or keeps running into a context limit. The more often one prompt goes out, the more a shorter version pays back.
| Industry | Who uses it | How they use it | Example |
|---|---|---|---|
| Software | Developers with long system prompts | Trim an agent’s standing instructions before it ships | A developer runs a long system prompt through it, so every later call sends fewer input tokens |
| Customer support | Support leads and their teams | Shorten the template the team pastes into each ticket reply | A team lead trims the reply template, then tries the short one on a few old tickets before swapping it in |
| Data and research | Analysts whose scripts send a prompt with every row | Cut the instruction that repeats on each row | Someone whose script summarizes spreadsheet rows shortens the line that goes out with each one |
| Marketing agencies | Writers who share prompt templates | Keep a library of client prompts short | An agency trims the brief template its writers reuse for every client post |
If you only ask short one-off questions, skip it.
In any AI chat, before a long prompt goes to a paid model. It earns its keep on prompts you send again and again.
| The prompt | Why a shorter one pays | What to check afterwards |
|---|---|---|
| An agent’s system prompt | It goes out with every task the agent runs | Each rule and output format is still there |
| The line a script sends per row | It repeats once for every row | The format your script reads back has not changed |
| A team’s reply template | Everyone pastes it into each ticket | The tone still fits your customers |
| Your own long request | The padding shows which words the model never needed | Names, numbers and limits came through |
What you need first:
Paste the skill’s prompt into the chat and put the prompt you want trimmed where it says {{input}}.
<the skill’s prompt>
<paste your prompt here, in place of {{input}}>A polite request for a summary, pasted in place of {{input}}.
Hello, I hope you are well. If it is no trouble, could you please write a short summary of the article below, under 100 words, and mention the author and the date. Thank you so much.
Back comes something in this shape, not a saved run. The word counts are the model’s own estimate.
Summarize the article below in under 100 words. Name its author and date. About 36 words before, 13 after.
Every requirement made it through: the limit of 100 words, the author and the date. Only the courtesy around them was cut.
If you are new. An AI model reads every word you send, and paid models charge for those words. Prompt diet keeps your instructions and drops the padding around them. Try it first on a request you sent an AI last week. Then check that the short version still asks for everything.
For experts. It is one prompt template run by whichever model the chat uses, so what counts as filler is that model’s call. Constraints, names, numbers and formats are kept by instruction, not by parsing. Cutting words can also cut the repetition some models lean on for tone, or the few-shot examples that anchor a format. The word counts come from the same model, not a tokenizer. Test the short prompt on your own evals before you swap it in.
Sokheng · Kit
Reads Khmer and English text from documents, receipts and photos of signs, using a detector for text lines and a recogniser for each line. Suits developers who…
Qwen · Kit
Qwen2.5-7B-Instruct as GGUF files in eight quantization sizes, to chat with locally through llama.cpp. For people who want a capable 7.6B model on their own co…
ACE-Step · Kit
Generates music from text with a 4B diffusion decoder, 50 steps and guidance control, aiming for higher audio quality than the base. For creators with a GPU of…
alirezarezvani · Skill
Adds 411 ready-made skills for Claude Code and other coding agents, grouped by role from engineering and product to marketing, finance and compliance. Installs…
block · Kit
Self-hostable workspace where people and AI agents share the same rooms, built on a Nostr relay that logs every action as a signed event. For teams that want t…
backnotprop · Kit
Opens your AI coding agent's plans, specs and code changes in a browser so you can annotate or comment on them and send the feedback straight back to the agent.