编程

less-token

试用

Compress verbose summary prompts into structured one-line instructions. Text-to-text translator only — no CLI, no API key, no install, no external dependencies. I-Lang has been tested across ChatGPT, Claude, Gemini, DeepSeek, Kimi, Qwen and GLM. Instruction-only, zero dependencies.

它能做什么

Save 40-65% tokens on summarization tasks. Compress verbose natural language prompts into structured one-line instructions that any AI understands.

技能文档

Less Token

Save 40-65% tokens on summarization tasks. Compress verbose natural language prompts into structured one-line instructions that any AI understands.

This skill is a text-to-text translator only. It does not access files, fetch URLs, execute commands, or call external services. It only converts your summarization prompts into compressed syntax.

What You Get

  1. 40-65% fewer tokens — Compress long summarization prompts into one-line instructions.
  2. Same result — AI produces identical output from the compressed instruction.
  3. Cross-platform — Compressed instructions work on ChatGPT, Claude, Gemini, DeepSeek, Kimi, 豆包, 元宝.
  4. No install — No CLI, no brew, no npm, no binary, no API key. Copy, paste, done.

How to Use

  1. Copy the full protocol text from this skill page
  2. Paste it into any AI conversation
  3. AI responds — ready to compress

Quick Test

After pasting, try:

  • "Compress this: Please summarize the key points from this document in 3 professional bullet points"
  • AI returns: [SUM|sty=bullets,cnt=3,ton=pro]=>[OUT]
  • 70% fewer tokens. Same result.

Compression Templates

What you wantVerbose promptCompressed
Short summary"Give me a brief summary of the main points"[SUM|len=short]=>[OUT]
3 bullet points"Summarize in 3 concise bullet points"[SUM|sty=bullets,cnt=3]=>[OUT]
Professional report"Create a professional executive summary in Markdown"[SUM|ton=pro,sty=executive,fmt=md]=>[OUT]
Key findings only"Extract only the key findings and important data"[SUM|key=findings]=>[OUT]
Summarize + translate"Summarize then translate to Chinese"[SUM|len=short]=>[TRANSLATE|lang=zh]=>[OUT]
Compare + summarize"Compare these two and summarize the differences"[CMP]=>[DIFF]=>[SUM|sty=bullets]=>[OUT]
Reformat summary"Summarize as bullet points in Markdown"[SUM|sty=bullets]=>[FMT|fmt=md]=>[OUT]

Before & After

Before (28 words):

Please read through this document carefully, identify the most important points and key takeaways, then write a concise professional summary using bullet points.

After (7 words):

[SUM|key=important,sty=bullets,ton=pro]=>[OUT]

75% fewer tokens. Same result.

Before (22 words):

Take the main findings from the text above and rewrite them as a short executive summary suitable for a business audience.

After (5 words):

[SUM|sty=executive,ton=pro]=>[OUT]

77% fewer tokens. Same result.

Comparison

FeatureCLI-based toolsLess Token
Install requiredYes (brew, npm, binary)No
API key requiredYesNo
Works onSingle platformAny AI platform
Token efficiencyStandard prompts40-65% fewer tokens
Setup time5-10 minutes30 seconds
External dependenciesMultipleZero

Tested Platforms

ChatGPT ✅ · Claude ✅ · Gemini ✅ · DeepSeek ✅ · Kimi ✅ · 豆包 ✅ · 元宝 ✅

License

MIT — Free to use, share, and build on.

© 2026 I-Lang Research, Eastsoft Inc., Canada.

相关技能

以 AI 机器人身份加入视频会议,提供语音、虚拟形象与屏幕共享四种模式。

作者 johnpatternai21 次安装8 星标

把自然语言描述转为结构化 JSON,并由 mcp-diagram-generator MCP 服务生成 Draw.io、Mermaid 或 Excalidraw 图表文件。

作者 nssa.io1.0k 次安装47 星标

按用户明确指令,在得到大脑(Get笔记)中保存、搜索并管理笔记与知识库。

作者 iswalle763 次安装66 星标

在本地磁盘以分类纯 Markdown 文件保存需要长期留存的事实,与智能体内置记忆并存。

作者 Iván1 次安装

从 AdMapix API 拉取广告创意、应用、榜单和收入预估等数据,原样返回结构化 JSON。

作者 fly0pants

诊断生产力系统反复失效的根因,给出最小干预——容量测算、瓶颈定位、可靠的本地记录。

作者 Iván2 次安装

adsorgcn 的更多技能

浏览全部技能

Compress natural language prompts into I-Lang — AI-native structured instructions. Output is text notation only — review before passing to execution agents.

作者 adsorgcn35 次安装2 星标

Stop learning prompt engineering. Tell AI what you want in plain language — AI writes a structured instruction for you in I-Lang. Copy it to other AIs as a well-structured starting point. Zero prompt skills needed. Generates text instructions only, no code, no install, no credentials. Results may vary by model.

作者 adsorgcn24 次安装

Lazarus v2 — Bring dead websites back to life. Keyword-in: hunt candidate dead domains with evidence. Domain-in: recover only pages with heuristic indexing evidence, pass a hard review gate, ship a machine-readable Recovery Bundle. 捡尸复活死站:热词挖坟、启发式收录分层、强制审查闸门、标准化产物包。默认改写素材不复制表达。

作者 adsorgcn8 次安装2 星标

DeAI — Improve AI-drafted text to sound naturally human. Three-layer editing: remove overused filler phrases, restructure for natural rhythm, mark positions for authentic personal voice. Adds review markers ([💬] [📝] [📊]) for user to fill in — does not generate content. Supports Chinese, English, Japanese, Korean.

作者 adsorgcn10 次安装2 星标

利用Google官方API实现全链路出海SEO自动化,批量生成PSEO页面,自动发布和GA4数据驱动复盘。

作者 adsorgcn8 次安装2 星标

微信公众号写作助手。200+篇实战验证的爆文结构引擎。你投喂素材,它输出MD文件+封面图提示词+自查报告。内置17条写作基因、品牌简称、平台合规、10项自查清单。不接受无素材请求,不是内容生成器。

作者 adsorgcn13 次安装3 星标