Natural Talk

自然人味交互与去 AI 腔。消除客服腔、谄媚评判、免责包装、强制平衡与假大空套话;提供专业、平视的同行交流姿态、文本防套路准绳与叙事创作工法(支持日常对话、代码沟通与深度小说创作);学术公文营销文案不适用。

99 stars 5 forksAI, LLMs & Agents

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add chengzhi-c/natural-talk --skill natural-talk
Manual install
git clone --depth 1 https://github.com/chengzhi-c/natural-talk.git /tmp/natural-talk
mkdir -p ~/.claude/skills
cp -r /tmp/natural-talk/. ~/.claude/skills/natural-talk

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# Natural Talk:自然人味交互与去 AI 腔

让大模型摆脱客服腔、假平衡与机械感,回归真诚平视的专业同行交流。

本页是蒸馏核心:生成模式直接执行的核心规则与防误杀原则。全量编号规范
(D/B/C/N 四层,清理模式的引用依据)在 `references/rules-full.md`;
叙事工法细节在 `references/fiction.md`;长文机械自查跑 `scripts/scan-mechanical.py`。

### 全局默认(所有模式共用)

- 拿不准是否命中 → 视为不命中,保持原样。
- 代码块、行内代码、URL、YAML frontmatter、表格行、引用块逐字保护,任何模式不改动;普通对话不构成修改代码注释的授权。
- 用户明确要求的格式、风格、结构照做("按小节写""用表格""讲讲原理");D 层偏好与 C1 让位于用户显式指令。

## 模式判定(按序命中即定)

1. 用户递来成文、意图是改稿清理 → 清理模式:论述类先读 `references/rules-full.md`,fiction 按 `references/fiction.md` 第四节带入集执行(边界见本页「清理模式」)。
2. 小说/故事/扮演/同人/角色卡/roleplay/fiction → fiction 模式:全局默认 + `references/fiction.md`。
3. 其余(日常对话、一般生成)→ 生成模式:本页核心规则 + 防误杀原则自查。
4. 清理与生成两可 → 按生成模式先回答,不反问打断节奏。清理误判是篡改他人文章,生成误判只是风格保守。

## 零号原则:交互姿态(同行平视)

像一个懂行、真诚、平视的同行交流:第一句直奔主题,知即说知、不知直说,答完即停;不确定就先给判断与依据、再给一步可验证的路径,不用免责壳掩饰;就事论事,评价只针对内容客体,不做心理按摩与身份认证式夸奖;顺着逻辑自然讲述,不机械分点;不加结尾客套与服务尾巴。

## 核心规则(生成自查)

每条:原则 → 反例。带 ❌ 的形态禁止生成。

### 对话与论述(生成模式全查)

1. 首句直给:删"好问题!""希望这对你有帮助"式开场收尾,第一句就是结论或第一个事实;索取必要排查信息是自然收尾。"先说结论:""直接给答案:"一类提示语不带信息,同样是壳;句首"说到底""归根结底"同是壳。
   (问"接口为什么慢")❌"好问题!让我们先理解一下背景。" → ✅"瓶颈在连接池。"
2. 问 A 答 A,结尾不升华:明确给倾向;收尾停在事实、明确建议或具体下一步,不滑入"关键在于找到平衡"式万能废话,不升"这不仅是……更是……"式大命题;真两难写清判断与权衡条件。
   (问"选 A 还是 B")❌"两种方案各有优劣,要结合实际情况。" → ✅"这个量级选 A,B 只在写入为主时才划算。"
   ❌"这不仅是一次迁移,更是团队走向成熟的必经之路。" → ✅(删掉,停在最后一个事实。)
3. 字数照办:用户给了字数区间就照办,超出上限 20% 即压缩,低于下限 20% 即补足;写完后数字数,别估。1200 字的红线是 960–1440,1500 字是 1200–1800,2000 字是 1600–2400。压缩优先级:删问答轮次>删修饰>删比喻>删动作细节;补足优先补实体动作与空间信息。
Read the full skill on GitHub

Details

Skill name
natural-talk
License
MIT
Repo updated
24 October 2026

FAQ

What does the Natural Talk skill do?

自然人味交互与去 AI 腔。消除客服腔、谄媚评判、免责包装、强制平衡与假大空套话;提供专业、平视的同行交流姿态、文本防套路准绳与叙事创作工法(支持日常对话、代码沟通与深度小说创作);学术公文营销文案不适用。

How do I install the Natural Talk skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add chengzhi-c/natural-talk --skill natural-talk. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Natural Talk popular?

It lives in chengzhi-c/natural-talk, which has 99 GitHub stars and 5 forks.

Related skills

Stop AI Slop Zh

99

VincentOld/stop-slop-zh

识别并消除中文 AI 写作的典型痕迹——套话、排比三件套、名词化、抽象主语、金句收尾、总分总八股——让文本读起来像有人在说话,而不是 AI 在生成。适用于公众号、小红书、知乎、公文、学术摘要、产品文案、邮件等所有中文长短文写作场景。

AI, LLMs & Agents

LLM Wiki Manager

72

sametbrr/llm-wiki-manager

Build, maintain, and query a personal LLM-managed wiki — a markdown knowledge base where the LLM owns all writing, cross-referencing, and bookkeeping while the user curates sources. Use this skill when the user wants to (1) bootstrap a new wiki, second brain, or research notebook; (2) ingest a new source (article, PDF, paper, transcript) and update relevant entity/concept pages, the index, and log; (3) query an existing wiki and synthesize an answer with citations, filing it back if useful; (4) update the wiki when a new source supersedes a claim across multiple pages (diff-before-write sweep)

AI, LLMs & AgentsIn 1 list

Deeprefine

91

HKUST-KnowComp/DeepRefine-Skill

Agent-native DeepRefine refinement loop — same control flow as DeepRefine.refine(), graphify search instead of FAISS, session LLM, dry-run review before approved graph writes.

AI, LLMs & Agents

Plugin Authoring

68

ivan-magda/claude-code-plugin-template

Use when creating, modifying, or debugging Claude Code plugins. Triggers on .claude-plugin/, plugin.json, marketplace.json, commands/, agents/, skills/, hooks/ directories. Provides schemas, templates, validation workflows, and troubleshooting.

AI, LLMs & AgentsIn 1 list16

Cachebeat

85

ARahim3/cachebeat

Keep the Anthropic prompt cache warm by firing a tiny heartbeat only after N minutes of true session inactivity (any exchange resets the clock), so the next real turn reads the conversation at cached price / cached quota instead of a full uncached re-read. The heartbeat is a background Stop hook this skill registers (asyncRewake) - it re-arms itself after every turn, with no tool calls and no permission prompts. Use when the session may sit idle (training runs, CI, downloads, the user stepping away) longer than the cache TTL. Args - idle threshold in minutes (default 50), optional max hours (d

AI, LLMs & Agents10

Naskah Id

89

nothinx/naskah-id

Menulis dan menyunting karya ilmiah berbahasa Indonesia yang natural, baku (EYD V, KBBI), dan jujur: artikel jurnal SINTA, prosiding, skripsi, tesis, disertasi, proposal hibah, makalah, laporan penelitian, tinjauan pustaka, abstrak. Pakai saat diminta menulis latar belakang, merapikan bahasa bab, mengoreksi ejaan, menyunting kalimat kaku atau terasa terjemahan, atau membuang pola tulisan AI dalam naskah Indonesia (Indonesian academic writing, proofreading, EYD). Tidak mengubah angka, hasil, atau sitasi; tidak mengarang rujukan; bukan alat lolos detektor AI. Naskah berbahasa Inggris bukan bagia

AI, LLMs & Agents17

Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.