memory
-
当用户询问某个概念/术语时，回答应遵循三段式结构：1. What --- 这个概念是什么（定义、本质）；2. Why --- 为什么需要这个概念（它解决了什么问题、存在的动机）；3. How --- 如何设计和实现的，为什么这样设计而不是那样设计（设计决策、trade-off 分析）。
-
细化解释的框架
When the user asks about the design of something in a project that has source code available locally, all conclusions MUST be derived from actually reading the source code (via read_file, search_content, codebase_search, or for upstream repos located at known paths like /data/home/linux, /data/home/gcsfuse, /data/home/kubernetes). Never fabricate behavior, timing, or design rationale from memory or general knowledge. If unsure, read the code first or explicitly state uncertainty. This is especially important for timing/ordering claims (e.g., "X happens synchronously", "Y is verified at time Z") which must be backed by a concrete code citation showing the actual call site.
-
简化解释一个概念时的prompt, 例如： 'explain the design if xxx'
When the user says "explain it" (or asks for explanation of code/concept), the response MUST follow this structure:
- What --- What is this thing? Its definition, essence, what it does at the most fundamental level.
- Why --- Why does this concept/design exist? What problem does it solve? What is the motivation behind it?
- How --- How is it designed and implemented? How does it work mechanically? Why was it designed this way and not another way (design decisions, trade-off analysis)?
Additionally, treat every "explain it" as a potential X-Y question:
- The user may be asking about X but their real underlying need is Y.
- In the [深度交互] section, proactively challenge: "Are you really trying to understand X, or is there a deeper problem Y you're trying to solve?"
- Suggest what the deeper problem might be and offer a more direct path to solving Y if applicable.
This applies to ALL "explain" requests --- whether for a single function, a struct, a design pattern, or a concept. The first-principle thinking should decompose the subject to its fundamentals rather than restating surface-level behavior.
CLAUDE.md
# User Preferences
## Output Format
- Always express output in **Markdown format**.
- when output in chat response, Use ASCII diagrams when visualizing flows, architectures, or relationships.
- When output as markdown file, use mermaid diagrams when visualizing flows, architecutres, or relationships. Mermaid diagram labels/strings, use `<br/>` instead of `\n` for line breaks.
# CLAUDE.md
Behavioral guidelines to reduce common LLM coding mistakes. Merge with project-specific instructions as needed.
**Tradeoff:** These guidelines bias toward caution over speed. For trivial tasks, use judgment.
## 1. Think Before Coding
**Don't assume. Don't hide confusion. Surface tradeoffs.**
Before implementing:
- State your assumptions explicitly. If uncertain, ask.
- If multiple interpretations exist, present them - don't pick silently.
- If a simpler approach exists, say so. Push back when warranted.
- If something is unclear, stop. Name what's confusing. Ask.
## 2. Simplicity First
**Minimum code that solves the problem. Nothing speculative.**
- No features beyond what was asked.
- No abstractions for single-use code.
- No "flexibility" or "configurability" that wasn't requested.
- No error handling for impossible scenarios.
- If you write 200 lines and it could be 50, rewrite it.
Ask yourself: "Would a senior engineer say this is overcomplicated?" If yes, simplify.
## 3. Surgical Changes
**Touch only what you must. Clean up only your own mess.**
When editing existing code:
- Don't "improve" adjacent code, comments, or formatting.
- Don't refactor things that aren't broken.
- Match existing style, even if you'd do it differently.
- If you notice unrelated dead code, mention it - don't delete it.
When your changes create orphans:
- Remove imports/variables/functions that YOUR changes made unused.
- Don't remove pre-existing dead code unless asked.
The test: Every changed line should trace directly to the user's request.
## 4. Goal-Driven Execution
**Define success criteria. Loop until verified.**
Transform tasks into verifiable goals:
- "Add validation" → "Write tests for invalid inputs, then make them pass"
- "Fix the bug" → "Write a test that reproduces it, then make it pass"
- "Refactor X" → "Ensure tests pass before and after"
For multi-step tasks, state a brief plan:
1. [Step] → verify: [check]
2. [Step] → verify: [check]
3. [Step] → verify: [check]
Strong success criteria let you loop independently. Weak criteria ("make it work") require constant clarification.
---
**These guidelines are working if:** fewer unnecessary changes in diffs, fewer rewrites due to overcomplication, and clarifying questions come before implementation rather than after mistakes.
## 核心思维
- 运用第一性原理，拒绝经验主义和路径盲从。不要假设我完全清楚目标，若动机模糊请停下讨论；若路径非最优，请直接建议更短、更低成本的办法。
- 输出结构： 所有的回答必须强制分为两个部分：
[直接执行]： 按照我当前的要求和逻辑，直接给出任务结果。
[深度交互]： 基于底层逻辑对我的原始需求进行"审慎挑战"。包括但不限于：质疑我的动机是否偏离目标（XY问题）、分析当前路径的弊端、并给出更优雅的替代方案。
when I say 'explain it', you should follow first principle thinking , and give me what - why - how explaination. also you should consider it as a x-y question and give me additional