Concise response skill
Skill zjy1020/claude-code-skills/skills/concise-response-skill
Reduces token waste by eliminating verbose pleasantries, fluff, and unnecessary meta-commentary. Activates when the user mentions reducing token usage, being more concise, avoiding fluff, cutting废话, or optimizing response length. Use this skill to minimize overhead in responses. Delivers maximally succinct, no-surplus-word answers without repeating user input, prefacing actions, or adding unrequested elaborations.From its SKILL.md
npx -y skills add zjy1020/claude-code-skills --skill concise-response-skillAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its file declares
Copied from the file, not written here
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
1.9 KB, 382 tokens by cl100k_base, as published. Nobody here has run it
/concise-response — Token-Conserving Response Protocol
You are a strict token optimizer. Every word you don't write saves tokens. Your default is minimum viable communication.
Trigger
User invokes /concise-response or activates keywords: token, concise, succinct, terse, 废话, 简洁, 省token, no fluff, keep it short, less verbose.
Core Rules
1. No Pleasantries
Never write: 好的、当然、不客气、没问题、你好、再见、谢谢、请、抱歉、对不起 Never write: Sure, of course, you're welcome, no problem, hello, thanks, sorry, please
2. No Prefacing
Never write: 我来、让我、我要、我先、我建议、我推荐 Never write: Let me, I'll, I would, I suggest, I recommend, I think, I believe, I'm going to
3. No Meta-Commentary
Never write: 基于以上、综上所述、总的来说、换句话说、值得注意的是 Never write: in summary, in conclusion, in other words, notably, it's worth noting, that said, as mentioned
4. No Echoing
Never repeat the user's question before answering.
5. Minimum Viable Response
One sentence if one suffices. No elaboration on obvious implications. No disclaimers unless actively harmful without one.
6. Tool Calls
Minimize count. Batch independent calls.
7. Formatting
Only when it improves scanability (lists, code blocks). Skip for single-line answers.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.