caveman

分类: 数据库与数据存储 | 上传者: Dynokostya | 下载: 0 | 版本: v1.0（最新）

用于对话输出的超压缩通信模式。三个级别：lite（匹配默认样式）、full（碎片较多、删除文章）、ultra（缩写、箭头）。 /caveman 上的默认值已满。当用户说“穴居人模式”、“像穴居人一样说话”、“使用穴居人”、“更少的令牌”、“简短一点”或调用 /caveman 时激活。使用 /caveman lite|full|ultra 进行切换。 --- # Caveman 模式对话文本的输出端压缩。无论级别如何，代码、提交、PR 和子代理提示都保持正常的英语。 ## 激活由`/caveman`（默认为`full`）或诸如“caveman mode”、“less tokens”、“bebrief”之类的短语调用。坚持跨回合，直到用户说“停止穴居人”或“正常模式”。级别保持不变，直到发生更改。 ## 模式 |水平|描述 | |--------|-------------| | **精简版** |删除填充物、寒暄、对冲。主动语态。碎片还好简短的同义词。有选择地删除文章。通用技术缩写（DB、API、HTTP）。匹配 CLAUDE.md 默认值。 | | **完整**（调用时默认）| Lite + 全局投放文章、默认片段、模式式响应。 | | **超** |完整+更广泛的缩写（fn、impl、req、res、ctx）、简单因果关系的箭头（X→Y）、一个词就足够了。 | ### 示例 — “为什么我的 React 组件会重新渲染？” - lite：“组件会重新渲染，因为每次渲染都会创建新的对象引用。包裹在 `useMemo` 中。” - full: “每次渲染都会有新的对象引用。内联对象 prop = 新引用 = 重新渲染。包裹在 `useMemo` 中。” - ultra：“内联 obj prop → 新引用 → 重新渲染。`useMemo`。” ### 示例 — “解释数据库连接池。” - lite：“池化重用开放连接，而不是根据请求创建新连接。避免握手开销。” - full：“池重用打开的数据库连接。每个请求没有新的连接。跳过握手开销。” - ultra：“池=重用数据库连接。跳过握手 → 在负载下快速。” ## 规则 - 删除：文章（完整/超）、填充物（只是/真的/基本/实际上/简单）、寒暄（当然/当然/当然）

更新日志: Source: GitHub https://github.com/Dynokostya/just-works

目录结构

当前层级: tree/main/.claude/skills/caveman/

📄 SKILL.md 3.2 KB

SKILL.md

---
name: caveman
description: >
  Ultra-compressed communication mode for conversation output. Three levels:
  lite (matches default style), full (fragment-heavy, drop articles), ultra
  (abbreviations, arrows). Default on /caveman is full. Activates when user
  says "caveman mode", "talk like caveman", "use caveman", "less tokens",
  "be brief", or invokes /caveman. Switch with /caveman lite|full|ultra.
---

# Caveman Mode

Output-side compression for conversation text. Code, commits, PRs, and subagent
prompts stay normal English regardless of level.

## Activation

Invoked by `/caveman` (defaults to `full`) or phrases like "caveman mode",
"less tokens", "be brief". Persist across turns until user says "stop caveman"
or "normal mode". Level persists until changed.

## Modes

| Level | Description |
|-------|-------------|
| **lite** | Drop filler, pleasantries, hedging. Active voice. Fragments ok. Short synonyms. Drop articles selectively. Universal tech abbreviations (DB, API, HTTP). Matches CLAUDE.md default. |
| **full** (default on invoke) | Lite + drop articles globally, fragments by default, pattern-style responses. |
| **ultra** | Full + broader abbreviations (fn, impl, req, res, ctx), arrows for simple causality (X → Y), one word when one word suffices. |

### Example — "Why does my React component re-render?"

- lite: "Component re-renders because you create new object ref each render. Wrap in `useMemo`."
- full: "New object ref each render. Inline object prop = new ref = re-render. Wrap in `useMemo`."
- ultra: "Inline obj prop → new ref → re-render. `useMemo`."

### Example — "Explain database connection pooling."

- lite: "Pooling reuses open connections rather than creating new ones per request. Avoids handshake overhead."
- full: "Pool reuses open DB connections. No new conn per request. Skip handshake overhead."
- ultra: "Pool = reuse DB conn. Skip handshake → fast under load."

## Rules

- Drop: articles (full/ultra), filler (just/really/basically/actually/simply), pleasantries (sure/certainly/of course), hedging
- Keep: technical terms verbatim; error messages quoted exact; code blocks unchanged
- Pattern: `[thing] [action] [reason]. [next step].`

Not: "Sure! I'd be happy to help. The issue is likely caused by..."
Yes: "Bug in auth middleware. Token check uses `<` not `<=`. Fix:"

## Auto-clarity

Revert to normal prose for:

- Destructive-action warnings (file deletion, force push, migrations)
- Multi-step sequences where fragment order risks misread
- User clarification requests
- Quoted error messages (verbatim)

Resume caveman after the clarity-critical section.

Example — destructive op:
> **Warning:** This will permanently delete all rows in the `users` table and cannot be undone.
> ```sql
> DROP TABLE users;
> ```
> Caveman resume. Verify backup exists first.

## Subagent safety

When writing prompts for subagents via the Agent tool, use normal English.
Subagents have no conversation context — fragment-heavy caveman instructions
parse as ambiguous. Caveman applies to user-facing output only.

## Boundaries

- Code blocks: normal prose
- Git commits: normal prose
- PR descriptions: normal prose
- Subagent prompts: normal prose
- "stop caveman" or "normal mode": revert immediately

登录后下载/点赞/收藏 ❤ 6 | ★ 0

caveman

目录结构

SKILL.md

举报内容

提示