@mulou/skills-for-gpt
v1.0.0
Published
Concise-response skills (English + Chinese) that constrain LLM output: conclusion-first, plain language, no filler. For Claude Code / Codex agents
Maintainers
Readme
skills-for-gpt
约束 LLM 输出风格的 skill 集合。当前包含一对内容完全对齐的中英文 skill,用于让模型的回答精简、结论先行、可直接使用。
Skills that constrain LLM output style. Currently contains a pair of fully-aligned English/Chinese skills that make model answers concise, conclusion-first, and ready to use.
安装 / Install
通过 npm 下载 / Download via npm:
npm install @mulou/skills-for-gpt安装后把 SKILL.md 复制到你的 agent skills 目录 / After install, copy the SKILL.md files into your agent's skills folder:
# Claude Code (英文版 / English)
cp -r node_modules/@mulou/skills-for-gpt/concise-response ~/.claude/skills/
# Claude Code (中文版 / Chinese)
cp -r node_modules/@mulou/skills-for-gpt/concise-response-zh ~/.claude/skills/Codex 用户把目标改为 ~/.codex/skills/。不想装依赖,可用 npm pack @mulou/skills-for-gpt 下载 tarball,解压即得两个 SKILL.md。
Codex users: target ~/.codex/skills/ instead. To grab the files without a dependency, run npm pack @mulou/skills-for-gpt and unpack the tarball.
Skills
| Skill | 语言 / Language | 说明 / Description |
|---|---|---|
| concise-response | English | Concise, conclusion-first answering style |
| concise-response-zh | 中文 | 精简回答风格(与英文版内容一致) |
核心规则 / Core Rules
两个 skill 强制执行同一套规则 / Both skills enforce the same rules:
- 结论先行 — 第一句话直接给答案,再解释必要原因。 Conclusion first — the first sentence is the answer; reasons come after.
- 语言简单准确 — 不堆砌术语,术语首次出现时解释。 Plain, precise language — no jargon stacking; explain terms on first use.
- 简单问题 1~3 句话 — 不加背景、免责声明或"另外"段落。 Simple questions get 1–3 sentences — no padding.
- 解释用「结论 → 原因 → 怎么做」结构。 Explanations follow Conclusion → Reason → What to do.
- 不注水 — 不复述问题、不重复结论、不扩展无关背景。 No filler — don't restate the question, repeat the conclusion, or add unrelated background.
- 不知道就明说 — 信息不足时说"目前无法确定",并指出缺什么。 Admit uncertainty — say what exactly is missing instead of guessing.
- 报告文字可直接使用 — 只给最终正式文本,不加元说明。 Report text is paste-ready — final formal wording only.
- 给一个答案,不给菜单 — 除非用户要求,不给长列表、大表格或多套方案。 One answer, not a menu — no Plan A/B/C unless asked.
- 风险警告是精简的例外 — 涉及安全、数据或工程风险时加必要警告,但保持一两句话。 Risk warnings are the exception — necessary, but kept to a sentence or two.
每个 skill 还包含发送前自检清单和反面模式(禁止出现的典型写法),把风格要求变成可核对的动作。 Each skill also includes a pre-send checklist and anti-patterns section that turn the style rules into checkable actions.
使用方法 / Usage
将 skill 目录放入你的 agent 的 skills 目录(例如 Codex 的 ~/.codex/skills/,或 Claude Code 的 .claude/skills/)。每个 skill 是一个包含 SKILL.md 的目录,frontmatter 中的 description 描述了触发条件,供 agent 自动匹配加载。
Place a skill directory into your agent's skills folder (e.g. ~/.codex/skills/ for Codex, or .claude/skills/ for Claude Code). Each skill is a directory containing a SKILL.md; the frontmatter description states the trigger conditions for automatic loading.
按需选择语言版本:面向中文对话用 concise-response-zh,面向英文对话用 concise-response,也可以两个都装。
Pick the version matching your conversation language — or install both.
