claude-comstyle
starsforks
How to install claude-comstyle?
bsquang/claude-comstyle
$git clone https://github.com/bsquang/claude-comstyle> 🌐 Language: English | Tiếng Việt
README.md
| 1 | # Claude Prompt Styles for Devs |
| 2 | |
| 3 | > 🌐 **Language:** **English** | [Tiếng Việt](README_VN.md) |
| 4 | |
| 5 | You control how Claude talks to you. Same answer, different style — pick the one that fits how you think. |
| 6 | |
| 7 | > Tested on Claude. May work with other AI assistants — results not guaranteed. |
| 8 | |
| 9 | --- |
| 10 | |
| 11 | ## ⚡ Quick Start |
| 12 | |
| 13 | Pick a style from the table → copy the prompt → paste at the start of your chat. |
| 14 | |
| 15 | Want it permanent? Add to `~/.claude/CLAUDE.md` — active for every session. |
| 16 | |
| 17 | **Want even fewer tokens?** Append `terminal CLI style` to any prompt → strips all markdown → plain text output → ~20–30% additional savings on top of the style. |
| 18 | |
| 19 | --- |
| 20 | |
| 21 | ## 📊 Style Overview |
| 22 | |
| 23 | | Style | Output tokens | Best for | Not suitable for | | |
| 24 | |---|---|---|---|---| |
| 25 | | 🪖 Military | 🟢 65–75% fewer | When you need the fix, not the explanation | Thinking through architecture | [↗](#-military) | |
| 26 | | 🪨 Caveman | 🟢 65–75% fewer | Fast back-and-forth while actively coding | Writing docs or PR descriptions | [↗](#-caveman) | |
| 27 | | 🔍 Reality Check | 🟢 60–70% fewer | Evaluating if what you built is actually worth it | When you want validation, not truth | [↗](#-reality-check) | |
| 28 | | 📋 git log | 🟢 50–65% fewer | Step-by-step instructions you'll paste into a ticket | Understanding why something works | [↗](#-git-log-style) | |
| 29 | | ❓ Socratic | 🟢 50–70% fewer* | Understanding WHY, not just HOW | When you're on a deadline | [↗](#-socratic) | |
| 30 | | 📌 BLUF | 🟡 20–35% fewer | Choosing between two options (X vs Y) | Simple one-answer questions | [↗](#-bluf-bottom-line-up-front) | |
| 31 | | 🧙 Yoda | 🔴 ~0% (neutral) | Livening up a boring debugging session | Long multi-step explanations | [↗](#-yoda) | |
| 32 | | 🏴☠️ Pirate | 🔴 +5–15% more | Live demos, team screenshots, memes | Any real debugging work | [↗](#️-pirate) | |
| 33 | | 💾 80s Hacker | 🔴 +5–15% more | Screencasts where drama helps | Anything you'll copy into docs | [↗](#-80s-hacker-movie) | |
| 34 | | 👨 Dad Joke | 🔴 +10–20% more | Team Slack when you want groans + learning | Daily solo use | [↗](#-dad-joke) | |
| 35 | | 🦆 Rubber Duck | 🔴 0–+20% more | Learning a concept from zero | Quick reference lookups | [↗](#-rubber-duck) | |
| 36 | | 🔬 Feynman | 🔴 +20–40% more | Onboarding juniors or learning unfamiliar topics | Experienced devs on familiar ground | [↗](#-feynman) | |
| 37 | | 🧱 First Principles | 🔴 +20–30% more | Picking a tech stack or architecture | Simple factual questions | [↗](#-first-principles) | |
| 38 | |
| 39 | *Socratic saves tokens per response but may cost more total across multi-turn conversations. |
| 40 | |
| 41 | --- |
| 42 | |
| 43 | ## Token legend |
| 44 | |
| 45 | | Badge | Meaning | |
| 46 | |---|---| |
| 47 | | 🟢 | Fewer output tokens than default | |
| 48 | | 🟡 | Slightly fewer output tokens | |
| 49 | | 🔴 | More output tokens than default | |
| 50 | |
| 51 | --- |
| 52 | |
| 53 | ## ⚡ Productivity Styles |
| 54 | |
| 55 | ### 🪖 Military |
| 56 | Terse, direct, no fluff. Every word earns its place. |
| 57 | |
| 58 | **Token savings: 🟢 ~65–75% fewer output tokens** |
| 59 | |
| 60 | **Prompt:** |
| 61 | ``` |
| 62 | Military style. Direct. No preamble. No filler. Facts only. |
| 63 | Format: [problem] → [cause] → [fix]. |
| 64 | Code unchanged. Technical terms intact. |
| 65 | ``` |
| 66 | |
| 67 | **Before / After** |
| 68 | |
| 69 | | 🗣️ Default Claude | 🪖 Military | |
| 70 | |---|---| |
| 71 | | Sure! A 401 typically means there's an authentication issue. The most common cause is an expired or invalid token. Looking at your middleware, it seems the token expiry check might be slightly off — specifically, the comparison operator used when validating the `exp` field. You'll want to review that logic and potentially add a token refresh flow before making the request. | Token expired. Middleware checks `exp` with `<` not `<=`. Fix: change to `<=` or refresh token before request. | |
| 72 | | *(VI)* Claude ơi tại sao API của tôi bị lỗi 401 hoài vậy, middleware tôi viết rồi mà vẫn không qua được? | *(VI)* Token hết hạn. Middleware kiểm tra `exp` dùng `<` thay vì `<=`. Sửa: đổi thành `<=` hoặc refresh token trước khi gọi request. | |
| 73 | |
| 74 | **+ `terminal CLI style`** |
| 75 | ``` |
| 76 | token expired → exp check < not <= → change to <= or refresh before request |
| 77 | ``` |
| 78 | |
| 79 | **Pros:** |
| 80 | - Massive token reduction |
| 81 | - Instant to read, no scrolling |
| 82 | - Forces AI to prioritize the actual a |
People who liked this also viewed
6 picksType
- tunellm chat in text filePromptsJul 2026
- snippetsSupercharge your Snippets in VS Code — Manage your code snippets & AI prompts without quitting your editor.PromptsJul 2026
- prompt-optimizerAn AI prompt optimizer for writing better prompts and getting better AI results.PromptsJul 2026
- chatgpt_system_promptA collection of GPT system prompts and various prompt injection/leaking knowledge.PromptsJul 2026
- prompt-masterA Claude skill that writes the accurate prompts for any AI tool. Zero tokens or credits wasted. Full context and memory retentionPromptsJun 2026
- python-prompt-toolkitLibrary for building powerful interactive command line applications in PythonPromptsMay 2026
