ai-factory.evolve

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Evolve - Skill Self-Improvement

Evolve - 技能自我提升

Analyze project context, patches, and codebase to improve existing skills. Makes AI Factory smarter with every run.

分析项目上下文、补丁和代码库，以改进现有技能。每次运行都能让AI Factory变得更智能。

Core Idea

核心思路

patches (past mistakes) + project context + codebase patterns
    ↓
analyze recurring problems, tech-specific pitfalls, project conventions
    ↓
enhance skills with project-specific rules, guards, and patterns

patches (past mistakes) + project context + codebase patterns
    ↓
analyze recurring problems, tech-specific pitfalls, project conventions
    ↓
enhance skills with project-specific rules, guards, and patterns

Workflow

工作流程

Step 0: Load Context

步骤0：加载上下文

Read
.ai-factory/DESCRIPTION.md
to understand:

Tech stack
Architecture
Conventions

读取
.ai-factory/DESCRIPTION.md
以了解：

技术栈
架构
约定规范

Step 1: Collect Intelligence

步骤1：收集情报

1.1: Read all patches

Glob: .ai-factory/patches/*.md

Read every patch. For each one, extract:

Problem category (null-check, async, validation, types, API, DB, etc.)
Root cause pattern (what class of mistake was made)
Prevention rule (what should be done differently)
Tags

1.2: Aggregate patterns

Group patches by tags and categories. Identify:

Recurring problems — same tag appears 3+ times? This is a systemic issue
Tech-specific pitfalls — problems tied to the stack (e.g., React re-renders, Laravel N+1)
Missing guards — what checks/patterns could have prevented these bugs

1.3: Read codebase conventions

Scan the project for patterns:

Linter configs (
```
.eslintrc
```
,
```
phpstan.neon
```
,
```
ruff.toml
```
, etc.)
Existing test patterns (test file structure, assertions used)
Error handling patterns (try/catch style, error types)
Logging patterns (logger used, format, levels)
Import conventions, file structure

1.1：读取所有补丁

Glob: .ai-factory/patches/*.md

读取每个补丁，并为每个补丁提取：

问题类别（空值检查、异步处理、验证、类型、API、数据库等）
根本原因模式（属于哪类错误）
预防规则（应该如何改进）
标签

1.2：聚合模式

按标签和类别对补丁进行分组，识别：

重复出现的问题 — 同一标签出现3次以上？这是系统性问题
技术栈特定陷阱 — 与技术栈相关的问题（例如React重渲染、Laravel N+1查询）
缺失的防护机制 — 哪些检查/模式本可以预防这些bug

1.3：读取代码库约定

扫描项目以识别模式：

代码检查器配置（
```
.eslintrc
```
、
```
phpstan.neon
```
、
```
ruff.toml
```
等）
现有测试模式（测试文件结构、使用的断言方式）
错误处理模式（try/catch风格、错误类型）
日志记录模式（使用的日志工具、格式、级别）
导入规范、文件结构

Step 2: Read Current Skills

步骤2：读取当前技能

Determine which skills to evolve:

If
```
$ARGUMENTS
```
contains a specific skill name → evolve only that skill
If
```
$ARGUMENTS
```
is "all" or empty → evolve all installed skills

Read each target skill's SKILL.md:

Glob: {{skills_dir}}/*/SKILL.md

If skills are not installed yet (no

{{skills_dir}}/

), read from source:

Glob: skills/*/SKILL.md

确定要进化的技能：

如果
```
$ARGUMENTS
```
包含特定技能名称 → 仅进化该技能
如果
```
$ARGUMENTS
```
为 "all" 或为空 → 进化所有已安装的技能

读取每个目标技能的SKILL.md：

Glob: {{skills_dir}}/*/SKILL.md

如果技能尚未安装（无

{{skills_dir}}/

），则从源目录读取：

Glob: skills/*/SKILL.md

Step 3: Analyze Gaps

步骤3：分析差距

For each skill, identify what's missing based on collected intelligence:

3.1: Patch-driven gaps

Compare patch patterns against skill instructions:

Does
```
/ai-factory.fix
```
mention the most common error categories from patches? If not → add them
Does
```
/ai-factory.implement
```
warn about the pitfalls found in patches? If not → add guards
Does
```
/ai-factory.task
```
include logging/validation requirements for problem areas? If not → enhance
Does
```
/ai-factory.review
```
check for the patterns that caused bugs? If not → add checklist items

3.2: Tech-stack gaps

Compare project tech stack against skill instructions:

Skills reference generic patterns but project uses specific framework? → Add framework-specific guidance
Project uses TypeScript but skills show JS examples? → Update examples
Project uses specific ORM (Prisma, Eloquent)? → Add ORM-specific patterns

3.3: Convention gaps

Compare project conventions against skill instructions:

Project has specific error handling pattern? → Skills should enforce it
Project uses specific logger? → Skills should reference it
Project has specific file structure? → Skills should follow it

针对每个技能，基于收集到的情报识别缺失内容：

3.1：补丁驱动的差距

对比补丁模式与技能说明：

```
/ai-factory.fix
```
是否提及补丁中最常见的错误类别？如果没有 → 添加相关内容
```
/ai-factory.implement
```
是否针对补丁中发现的陷阱发出警告？如果没有 → 添加防护规则
```
/ai-factory.task
```
是否包含针对问题领域的日志/验证要求？如果没有 → 增强相关内容
```
/ai-factory.review
```
是否检查导致bug的模式？如果没有 → 添加检查清单项

3.2：技术栈差距

对比项目技术栈与技能说明：

技能引用通用模式，但项目使用特定框架？→ 添加框架特定指导
项目使用TypeScript，但技能展示JS示例？→ 更新示例
项目使用特定ORM（Prisma、Eloquent）？→ 添加ORM特定模式

3.3：约定规范差距

对比项目约定与技能说明：

项目有特定的错误处理模式？→ 技能应强制执行该模式
项目使用特定日志工具？→ 技能应引用该工具
项目有特定文件结构？→ 技能应遵循该结构

Step 4: Generate Improvements

步骤4：生成改进方案

For each gap found, create a concrete improvement:

markdown

undefined

针对每个发现的差距，创建具体的改进内容：

markdown

undefined

Improvement: [skill-name]

改进：[技能名称]

What

改进内容

[Specific change to make]

[具体要做的修改]

Why

原因

[Which patches/patterns drove this change]

[驱动此修改的补丁/模式]

Where

修改位置

[Exact section in SKILL.md to modify]

[SKILL.md中要修改的具体章节]

Change

修改内容

[The actual text to add/modify]


**Quality rules for improvements:**
- Each improvement must be traceable to a patch, convention, or tech-stack fact
- No generic advice — only project-specific enhancements
- Improvements must be minimal and focused — don't rewrite entire skills
- Preserve existing skill structure — add, don't replace

[要添加/修改的实际文本]


**改进内容的质量规则：**
- 每项改进必须可追溯到补丁、约定规范或技术栈事实
- 不提供通用建议 — 仅针对项目的特定增强
- 改进必须最小化且聚焦 — 不要重写整个技能
- 保留现有技能结构 — 仅添加内容，不替换

Step 5: Present & Apply

步骤5：展示与应用

5.1: Present improvements to user

undefined

5.1：向用户展示改进方案

undefined

Skill Evolution Report

技能进化报告

Based on:

X patches analyzed
Y recurring patterns found
Z tech-stack specific insights

基于：

分析了X个补丁
发现Y个重复模式
Z个技术栈特定洞察

Proposed Improvements

拟议改进

/ai-factory.fix

Add null-check guard — 5 patches involved null references → Add to Step 2: "Check for optional/nullable fields before accessing nested properties"
Add async/await pattern — 3 patches involved unhandled promises → Add to Important Rules: "Always use try/catch with async/await"

添加空值检查防护 — 5个补丁涉及空引用 → 添加到步骤2："在访问嵌套属性前检查可选/可空字段"
添加async/await模式 — 3个补丁涉及未处理的Promise → 添加到重要规则："始终将async/await与try/catch配合使用"

/ai-factory.implement

Add Prisma-specific warning — 2 patches from incorrect Prisma queries → Add to Logging: "Log all Prisma queries in DEBUG mode"

添加Prisma特定警告 — 2个补丁来自不正确的Prisma查询 → 添加到日志记录："在DEBUG模式下记录所有Prisma查询"

/ai-factory.review

Add checklist item — optional chaining not checked → Add to Correctness: "Optional chaining for nullable relations"

Apply these improvements?

Yes, apply all
Let me pick which ones
No, just save the report


**5.2: Apply approved improvements**

For each approved improvement:
1. Read the target SKILL.md
2. Apply the change using `Edit`
3. Keep changes minimal and surgical

**5.3: Save evolution log**

Create `.ai-factory/evolutions/YYYY-MM-DD-HH.mm.md`:

```bash
mkdir -p .ai-factory/evolutions

markdown

undefined

添加检查清单项 — 未检查可选链 → 添加到正确性检查："可空关联的可选链使用"

是否应用这些改进？

是，全部应用
让我选择要应用的项
否，仅保存报告


**5.2：应用已批准的改进**

针对每个已批准的改进：
1. 读取目标SKILL.md
2. 使用`Edit`应用修改
3. 保持修改最小化且精准

**5.3：保存进化日志**

创建 `.ai-factory/evolutions/YYYY-MM-DD-HH.mm.md`：

```bash
mkdir -p .ai-factory/evolutions

markdown

undefined

Evolution: YYYY-MM-DD HH:mm

进化记录：YYYY-MM-DD HH:mm

Intelligence Summary

情报摘要

Patches analyzed: X
Recurring patterns: [list]
Tech stack: [from DESCRIPTION.md]

分析的补丁数量：X
重复模式：[列表]
技术栈：[来自DESCRIPTION.md]

Improvements Applied

已应用的改进

[skill-name]

[技能名称]

[change description] ← driven by patches: [patch filenames]

[修改描述] ← 驱动来源：[补丁文件名]

[skill-name]

[技能名称]

[change description] ← driven by: [tech stack / convention]

[修改描述] ← 驱动来源：[技术栈 / 约定规范]

Patterns Identified

识别的模式

[pattern]: [frequency] occurrences
[pattern]: [frequency] occurrences

undefined

[模式]：出现[频次]次
[模式]：出现[频次]次

undefined

Step 6: Suggest Next Actions

步骤6：建议后续操作

undefined

undefined

Evolution Complete

进化完成

Skills improved: X Improvements applied: Y

已改进技能数量：X 已应用改进数量：Y

Recommendations

建议

Run
/ai-factory.review
on recent code to verify improvements
Next evolution — run
```
/ai-factory.evolve
```
again after 5-10 more fixes
Consider new skill — if pattern X keeps recurring, create a dedicated skill:
```
/ai-factory.skill-generator <skill-name>
```

undefined

运行
/ai-factory.review
检查近期代码，验证改进效果
下次进化 — 在完成5-10次更多修复后，再次运行
```
/ai-factory.evolve
```
考虑新增技能 — 如果模式X持续出现，创建专用技能：
```
/ai-factory.skill-generator <skill-name>
```

undefined

Context Cleanup

上下文清理

Context is heavy after reading all patches and skills. All improvements are saved — suggest freeing space:

AskUserQuestion: Free up context before continuing?

Options:
1. /clear — Full reset (recommended)
2. /compact — Compress history
3. Continue as is

读取所有补丁和技能后，上下文占用较大。所有改进已保存 — 建议释放空间：

AskUserQuestion: 在继续之前释放上下文空间？

选项：
1. /clear — 完全重置（推荐）
2. /compact — 压缩历史记录
3. 保持当前状态继续

What Each Skill Can Learn

各技能可学习内容

Skill	Learns From	Example Enhancement
`/ai-factory.fix`	Patches → common errors	"Check for X before accessing Y"
`/ai-factory.implement`	Patches → prevention rules	"When creating models, always validate Z"
`/ai-factory.task`	Patches → logging gaps	"Add validation task for nullable fields"
`/ai-factory.review`	Patches → missed issues	"Check: are all optional relations null-safe?"
`/ai-factory.commit`	Codebase → conventions	"Use project's commit prefix format"
`/ai-factory.feature`	Codebase → patterns	"Default branch prefix based on project convention"

技能	学习来源	示例改进
`/ai-factory.fix`	补丁 → 常见错误	"在访问Y之前检查X"
`/ai-factory.implement`	补丁 → 预防规则	"创建模型时，始终验证Z"
`/ai-factory.task`	补丁 → 日志缺口	"添加可空字段的验证任务"
`/ai-factory.review`	补丁 → 遗漏问题	"检查：所有可选关联是否都做了空值安全处理？"
`/ai-factory.commit`	代码库 → 约定规范	"使用项目的提交前缀格式"
`/ai-factory.feature`	代码库 → 模式	"基于项目约定的默认分支前缀"

Important Rules

重要规则

Traceable — every improvement must link to a patch, convention, or tech fact
Minimal — add rules, don't rewrite skills
Reversible — user approves before changes are applied
Cumulative — each evolution builds on previous ones
No hallucination — only suggest improvements backed by evidence
Preserve structure — don't change skill workflow, only enrich it

可追溯 — 每项改进必须关联到补丁、约定规范或技术事实
最小化 — 添加规则，不要重写技能
可逆 — 在应用修改前需获得用户批准
累积性 — 每次进化都基于之前的成果
无幻觉 — 仅建议有证据支持的改进
保留结构 — 不要改变技能工作流程，仅丰富内容

Examples

示例

Example 1: After 10 fixes with null-reference patterns

示例1：在10次修复后，空引用模式频繁出现

/ai-factory.evolve fix

→ Found 6/10 patches tagged #null-check
→ Improvement: Add to /ai-factory.fix Step 2:
  "PRIORITY CHECK: Look for optional/nullable fields accessed
   without null guards. This is the #1 source of bugs in this project."
→ Improvement: Add to /ai-factory.review checklist:
  "- [ ] All nullable DB fields have null checks in UI/API layer"

/ai-factory.evolve fix

→ 发现10个补丁中有6个标记为#null-check
→ 改进：添加到/ai-factory.fix步骤2：
  "优先级检查：在访问嵌套属性前，检查可选/可空字段。这是本项目中排名第一的bug来源。"
→ 改进：添加到/ai-factory.review检查清单：
  "- [ ] 所有可空数据库字段在UI/API层都有空值检查"

Example 2: Laravel project with N+1 issues

示例2：存在N+1问题的Laravel项目

/ai-factory.evolve all

→ Stack: Laravel + Eloquent (from DESCRIPTION.md)
→ Found 3 patches tagged #n-plus-one #database
→ Improvement: Add to /ai-factory.implement logging:
  "Enable query logging: DB::enableQueryLog() in DEBUG mode"
→ Improvement: Add to /ai-factory.review checklist:
  "- [ ] Eager loading used for related models (no N+1)"
→ Improvement: Add to /ai-factory.task descriptions:
  "Include 'use ->with() for relations' in DB-related tasks"

/ai-factory.evolve all

→ 技术栈：Laravel + Eloquent（来自DESCRIPTION.md）
→ 发现3个补丁标记为#n-plus-one #database
→ 改进：添加到/ai-factory.implement日志记录：
  "启用查询日志：在DEBUG模式下使用DB::enableQueryLog()"
→ 改进：添加到/ai-factory.review检查清单：
  "- [ ] 关联模型使用了预加载（无N+1问题）"
→ 改进：添加到/ai-factory.task描述：
  "在数据库相关任务中包含'使用->with()处理关联'"

Example 3: First run with no patches

示例3：首次运行，无补丁

/ai-factory.evolve

→ No patches found (first run)
→ Analyzing project context only...
→ Stack: Next.js 14 + Prisma + TypeScript
→ Improvement: Add to /ai-factory.implement:
  "Use server actions for mutations, API routes for external APIs"
→ Improvement: Add to /ai-factory.fix:
  "For Prisma errors, check schema.prisma for field types first"
→ Improvement: Add to /ai-factory.review:
  "- [ ] Server components don't use client-only hooks"

/ai-factory.evolve

→ 未找到补丁（首次运行）
→ 仅分析项目上下文...
→ 技术栈：Next.js 14 + Prisma + TypeScript
→ 改进：添加到/ai-factory.implement：
  "使用Server Actions处理突变，API路由处理外部API"
→ 改进：添加到/ai-factory.fix：
  "对于Prisma错误，首先检查schema.prisma中的字段类型"
→ 改进：添加到/ai-factory.review：
  "- [ ] 服务端组件未使用仅客户端钩子"

DO NOT:

禁止操作：

Do not rewrite entire skills
Do not remove existing rules
Do not add generic advice ("write clean code")
Do not create new skills (suggest using
```
/ai-factory.skill-generator
```
instead)
Do not apply changes without user approval
Do not evolve skills not installed in the project

不要重写整个技能
不要删除现有规则
不要添加通用建议（如"编写整洁代码"）
不要创建新技能（建议使用
```
/ai-factory.skill-generator
```
替代）
不要在未获得用户批准的情况下应用修改
不要进化项目中未安装的技能