About this skill

The Problem

Summarization, comparison, and translation tasks often rely on verbose natural-language prompts. Phrases such as “please carefully read”, “then write”, and “suitable for a business audience” add repeated context that increases token usage across repeated calls. Less Token focuses on compressing prompt text itself, not replacing the model, executing tools, or calling external services.

How It Works

This skill performs a text-to-text transformation: it rewrites long summarization prompts into one-line structured instructions, such as:

[SUM|sty=bullets,cnt=3]=>[OUT] requests three bullet points
[SUM|ton=pro,sty=executive,fmt=md]=>[OUT] requests a professional executive summary in Markdown
[SUM|len=short]=>[TRANSLATE|lang=zh]=>[OUT] requests a short summary followed by Chinese translation
[CMP]=>[DIFF]=>[SUM|sty=bullets]=>[OUT] requests comparison, difference extraction, then bullets

Usage is to paste the protocol into any AI conversation that understands it, then provide the original task. It claims compatibility with ChatGPT, Claude, Gemini, DeepSeek, Kimi, Doubao, and Yuanbao, with output expected to match the verbose prompt.

Boundaries

It does not access files, fetch URLs, execute commands, or call external APIs. Compression depends on the target model’s understanding of the syntax. For complex constraints, long-context reasoning, code execution, or permission-sensitive workflows, use explicit natural-language system prompts or structured tool definitions. It is best suited to high-volume summarization, key-point extraction, report rewriting, and translate-after-summary tasks.

Screenshots

Use Cases

CS lead turns long customer feedback into 3 key points for the daily service summary
Analyst compresses verbose research notes into a professional executive summary in Markdown
Ops team compares two competitor specs and extracts 3 key changes as bullets
Translation team summarizes an English source first, then requests a Chinese output

Best For

Prompt engineers maintaining reusable low-token instruction templates for batch summarization
Ops and product staff who regularly condense user feedback or competitor materials into bullets
Analysts who turn long text into executive summaries or key difference bullets
Multi-model users who need consistent summarization output across LLM chat platforms