Skip to content
Back to skills

Pdlc Perf

ASecurity

性能优化

  • 16 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added May 26, 2026
code-qualitybashtestinggitapi

Works with

  • terminal
  • api

Security analysis

A100/100

Scanned September 25, 2026

npx -y skills add kanfu-panda/pdlc-skills --skill pdlc-perf --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Pdlc Perf?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Pdlc Perf
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/kanfu-panda-pdlc-perf/badge)](https://www.skillsdirectory.com/skills/kanfu-panda-pdlc-perf)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: pdlc-perf
description: 性能优化
argument-hint: <功能ID | 优化目标>
allowed-tools: Read, Write, Edit, Glob, Grep, Bash
layer: 3
stage: quality
produces:
  - docs/04_testing/perf/<feature-id>-report.md
requires: []
next_step: null
terminal_state: null
---

# 性能优化

<!-- @include templates/prompts/iron-law.md(已内联于下方,无需另读) -->
⛔ **IRON LAW · 不可违反的硬门禁**

以下规则为**不可协商**的执行约束:

1. **文件必须落盘**:所有带编号(功能ID / 缺陷ID)的文档,必须作为实际文件写入磁盘,不可仅在对话中输出。
2. **阶段必须落章**:每个阶段完成后必须在状态机 `docs/.pdlc-state/<feature-id>.json` 追加 history,不可跳过。
3. **测试必须存在**:进入 `/pdlc-implement` 前,对应测试必须存在且处于红灯状态。违反则中止。
4. **自检必须执行**:段二自检为强制步骤,不得以"已经很好了"为由跳过。
5. **防循环**:段三修复为单次,不递归。无法自动修复的问题记录到报告,继续往下走。
6. **状态必推进**:成功执行某 phase 后 `current_stage` 必须变更。收尾时若发现 `current_stage` 未推进,视为失败并报错,**不得静默返回**(防止外层循环拿滞后的状态空转烧额度)。唯一例外:命中人工点主动 block 时,`current_stage` 保持不变但必须写 `last_phase_result.ok=false` + `blocked_reason`。

**违反任一条 = 立即中止当前命令,输出违规详情,等待人工介入。**
<!-- @include-end templates/prompts/iron-law.md -->

对指定服务或应用进行性能分析并提出优化方案。

## 分析维度

### 后端性能
- **数据库层**: 慢查询、N+1 问题、缺失索引、全表扫描
- **缓存层**: 缓存命中率、缓存策略(过期/淘汰)、缓存穿透/击穿/雪崩
- **接口层**: 响应时间、并发处理、连接池配置
- **代码层**: 算法复杂度、内存泄漏、不必要的序列化/反序列化

### 前端性能
- **加载性能**: 首屏时间、资源体积、代码分割、懒加载
- **运行时性能**: 不必要的重渲染、大列表虚拟滚动、防抖/节流
- **网络优化**: 请求合并、资源压缩、CDN 配置
- **缓存策略**: 浏览器缓存、Service Worker、本地存储

## 工作流程
1. **阅读代码**: 分析目标服务/应用的核心逻辑
2. **识别瓶颈**: 标注可能存在性能问题的代码段
3. **提出方案**: 对每个问题给出优化方案和预期收益
4. **实施优化**: 按优先级实施优化代码
5. **【必须创建文件】** 在 `docs/07_reviews/code/` 下创建性能优化报告

> ⚠️ **必须创建文件,不可仅在对话中输出。**

## 报告格式

文件名: `YYYYMMDD-<服务名>-perf-report.md`

**文档顶部包含 PDLC 追溯头**:
```
<!-- PDLC-TRACE -->
<!-- 功能名称: <服务名> -->
<!-- 阶段: 性能优化 -->
<!-- 创建时间: <ISO 8601> -->
```

**报告内容**:

| 序号 | 位置 | 问题描述 | 影响程度 | 优化方案 | 预期收益 |
|------|------|----------|----------|----------|----------|

**创建后验证**:确认文件已存在于 `docs/07_reviews/code/` 目录

## 要求
<!-- @include templates/prompts/output-language.md(已内联于下方,无需另读) -->
🌐 **Output language for generated artifacts**

All generated artifacts (PRDs, design docs, code comments, review reports,
test plans, deployment manuals, changelog entries, etc.) follow this policy:

1. **Default — match the conversation language exactly**:
   - 用户用中文与 Claude 对话 → 产中文文档、中文代码注释、中文报告
   - User talks to Claude in English → produce English artifacts
   - User talks in another language → produce artifacts in that language
   - **Never silently default to a fixed language regardless of the user's input.**

2. **Explicit override always wins**: when the user specifies a language for
   an artifact (e.g. "write the PRD in English", "用英文写 API 设计文档",
   "output the deploy doc in Japanese"), use that language for that artifact,
   regardless of conversation language.

3. **Mixed-language requirements**: if the user wants some artifacts in one
   language and others in a different language (common: Chinese PRD + English
   API docs for partners), honour each per-artifact instruction.

4. **Uncertain**: if you cannot reliably detect the conversation language,
   ask once before producing the first artifact.

This policy applies to **content** (prose, comments, headings). It does
**not** override technical conventions like English variable names, English
git commit subjects, or English error codes when the project's conventions
require them.
<!-- @include-end templates/prompts/output-language.md -->
- 优化方案按投入产出比排序(性价比高的优先)
- 给出优化前后的代码对比
- 不要为了优化而牺牲代码可读性

优化目标: $ARGUMENTS

<!-- @include templates/prompts/handoff.md(已内联于下方,无需另读) -->
## 段四:交接(Handoff)

命令完成后必须输出以下格式的最终消息:

```
✅ <阶段名> 完成:<主要产出物路径>
📊 自检:<通过数>/<总数> 通过(若有未通过,附要点)
📦 状态快照:docs/.pdlc-state/<feature-id>.json
👉 下一步:/pdlc-<next_step>
   (如果有分叉)或 /pdlc-<alt>(条件:<选择依据>)
```

**规则:**
- 主流程命令(写状态机的命令;下一跳见正文里「本命令的状态机取值」)必须显式输出"下一步",不可省略
- 工具型命令(Layer 3)可以没有 `next_step`,此时输出 `👉 下一步:(本次流程结束,无后续)`
- 分叉场景必须说明**选择条件**,例如"若需补充测试用例 → `/pdlc-tdd`;若测试已齐 → `/pdlc-review`"
<!-- @include-end templates/prompts/handoff.md -->

**本命令的 handoff 输出:**

```
✅ 性能优化报告 完成
📦 产出:docs/04_testing/perf/<feature-id>-report.md
👉 下一步:(本次流程结束,无后续)
```

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…