bobo-ai-cli 3.0.4 → 3.0.5

package/knowledge/llm-wiki-pattern.md

@@ -0,0 +1,71 @@
+# LLM-Wiki Pattern (Karpathy, 2026-04)
+
+Source: https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f
+
+## Core Insight
+
+RAG = 每次从头查原始文档，知识永远散装。
+Wiki = LLM 持续编译、互链、维护结构化知识，读的时候直接拿编译好的结果。
+
+**关键区别**: wiki 是持续积累的产物。交叉引用已经建好，矛盾已经标记，综合已经反映了你读过的所有东西。
+
+## 三层架构
+
+| 层 | 内容 | 权限 |
+|---|------|------|
+| Raw Sources | 原始文档（文章/论文/笔记/数据） | 只读不改 |
+| Wiki | LLM 生成的结构化页面（摘要/实体/概念/对比） | LLM 写+维护 |
+| Schema | wiki 约定+工作流规则 | 人+LLM 共同演进 |
+
+## 三个核心工作流
+
+### Ingest（新源进入）
+- 读源 → 讨论要点 → 写摘要页 → 更新 index → 更新相关实体/概念页 → 追加 log
+- 一个源通常触碰 10-15 个 wiki 页面
+- 可以逐个 ingest（参与度高）或批量 ingest（低监督）
+
+### Query（查询 + 回写）
+- 先读 index → 找相关页 → 合成回答（带引用）
+- **关键**: 好的答案应该回写为新 wiki 页（对比表/分析/发现不应消失在 chat 里）
+- 探索过程和 ingest 一样积累到知识库
+
+### Lint（定期健检）
+- 矛盾：不同页面对同一事实说法不一致
+- 过时：新源已更新但旧页面没跟上
+- 孤儿页：无入站链接
+- 缺页：被多次提及但没有独立页面的概念
+- 缺交叉引用：应该互链但没链的页面
+- 数据缺口：可以通过 web search 补充的空白
+
+## 两个关键文件
+
+### index.md（内容导向）
+- 所有页面的目录：链接 + 一句话摘要 + 元数据
+- 按类别组织（实体/概念/源/分析）
+- 每次 ingest 后更新
+- 查询时先读 index 再钻取
+
+### log.md（时序导向）
+- Append-only 操作记录
+- 格式: `## [YYYY-MM-DD] action | title`
+- 可 grep 解析: `grep "^## \[" log.md | tail -5`
+- 记录 ingest/query/lint 事件
+
+## 应用场景
+- 个人知识管理（目标/健康/自我提升）
+- 深度研究（论文/报告/weeks-long topic）
+- 读书笔记（角色/主题/情节线索互链）
+- 团队内部 wiki（Slack/会议/客户通话 → LLM 维护）
+- 竞品分析、尽职调查、旅行规划、课程笔记
+
+## 工具生态
+- Obsidian: 浏览 wiki、图谱视图、Web Clipper
+- qmd: 本地 Markdown 搜索（BM25 + 向量 + LLM re-rank）
+- Marp: Markdown → 幻灯片
+- Dataview: YAML frontmatter 查询
+- Git: 版本控制 + 协作
+
+## 核心洞察
+> 维护知识库的苦差事不是阅读或思考——是记账。更新交叉引用、保持摘要最新、标记新旧数据矛盾、维护一致性。人类放弃 wiki 是因为维护负担增长快于价值。LLM 不会无聊、不会忘记更新交叉引用、能一次触碰 15 个文件。
+
+> 人的工作是策展源、引导分析、提好问题、思考意义。LLM 的工作是其他一切。

package/knowledge/long-task-management.md

@@ -0,0 +1,79 @@
+# 长任务管理协议
+
+## 判定标准
+预计 ≥30 分钟的异步任务，必须走此协议。
+
+## 启动清单
+
+### 1. 设置心跳监控
+```
+启动后 10 分钟内必须设心跳：
+- 进程活着 + 有进展 + 资源健康
+- 完成后自动执行后续
+- 失败自愈（3 次失败才找人）
+```
+
+### 2. 三文件模式
+```
+task_plan.md  — 规划 + 进度追踪
+notes.md      — 过程笔记 + 发现
+[deliverable] — 最终交付物
+```
+
+### 3. Plans.md 表格（6+ 步任务）
+```markdown
+| Task | 内容 | DoD（Yes/No） | Depends | Status |
+|------|------|-------------|---------|--------|
+| T1   | ...  | ...         | -       | TODO   |
+```
+- Status: TODO → WIP → Done → Verified
+- Depends=`-` 可并行
+- 循环依赖 → 报错重规划
+
+## 进度汇报
+
+### 里程碑快照格式
+```
+📊 Progress: Task X/Y 完成
+状态: [在做/卡住/等决策]
+已完成: [列表]
+进展: [当前在干什么]
+ETA: [预估完成时间]
+需要: [如果需要对方做什么]
+```
+
+### 卡点上报时机
+- 重复失败 ≥3
+- 流程阻塞（等人/等资源）
+- 信息缺失
+- 工具/环境故障
+
+## Cron 心跳规范
+
+### timeout 设置
+- 纯脚本/curl: 60-300s
+- 涉及 LLM 多轮对话: **900s（15min）**
+- 涉及视频生成/训练: 1800s+
+
+### 自愈策略
+```
+失败 1 次 → 重试（不同策略）
+失败 2 次 → 重试（换方案）
+失败 3 次 → 上报 + 暂停
+```
+
+### 完成后自动后续
+- 下载产出
+- 跑验证
+- 通知相关方
+- 更新状态
+
+## 资源监控
+
+启动大任务前检查：
+- `df -h /` — 磁盘空间（<10% 就先清理）
+- `free -m` — 内存
+- 网络连通性
+- API quota/余额
+
+遇到 `No space left on device` → **1小时内同步卡点**，不自己死磕。

package/knowledge/memory/cache-optimization-and-skill-integration.md

@@ -1,102 +1,102 @@
----
-id: "cache-optimization-and-skill-integration"
-title: "缓存优化和技能整合 (2026-04-01)"
-type: "custom"
-tags: [memory, template]
-template: true
----
-
-# 缓存优化和技能整合 (2026-04-01)
-
-## cc-cache-fix 集成
-
-**状态**: ✅ 已验证和集成
-
-**核心发现**:
-- cc-cache-fix 解决 60-70% token 燃烧问题（相对基准）
-- 三个补丁：Delta 附件持久化、Hash 稳定性、TTL 扩展
-- 需要配合 context 压缩、会话分割、缓存预热才能达到最优效果
-
-**立即应用**:
-```bash
-git clone https://github.com/Rangizingo/cc-cache-fix.git
-cd cc-cache-fix && ./install.sh
-alias claude=claude-patched
-```
-
-**文档**: `rules/domain/cache-management.md`
-
----
-
-## 技能库整合计划
-
-**总体目标**: 47 个技能 → 21 个统一技能体系
-
-**分类**:
-- Agent 工程 (11) → 5 个合并技能
-- 搜索数据 (3) → 2 个合并技能
-- 创作设计 (4) → 3 个合并技能
-- 视频音频 (7) → 4 个合并技能
-- 浏览器爬虫 (3) → 1 个合并技能
-- 工具集成 (4) → 2 个合并技能
-- 安全运维 (3) → 1 个合并技能
-- 知识工具 (12) → 1 个独立仓库
-- 保持独立 (9) → 9 个独立技能
-
-**核心合并** (P0):
-1. context-optimization-suite (context-budget-analyzer + context-compressor)
-2. decision-making-framework (high-agency + structured-decision-alignment)
-3. memory-evolution-system (memory-manager + proactive-self-improving)
-4. quality-assurance-framework (adversarial-verification + self-rationalization-guard)
-5. cache-optimization-skill (cc-cache-fix 集成) ⭐ NEW
-
-**文档**: 
-- `skills/SKILL_INTEGRATION_MAP.md` - 技能整合地图
-- `skills/MIGRATION_GUIDE.md` - 迁移指南
-
----
-
-## 工程化规则更新
-
-**已更新**:
-- `rules/domain/cache-management.md` - 新增缓存管理规则
-- `rules/performance.md` - 添加缓存优化优先级
-
-**需要更新**:
-- `rules/domain/engineering-workflows.md` - 添加缓存预热步骤
-- `CLAUDE.md` - 添加技能体系说明
-- `capabilities/skills-guide.md` - 更新技能列表
-
----
-
-## 预期收益
-
-**短期** (1个月):
-- Token 节省: 60-70%
-- 技能数量: -30% (47 → 33)
-- 学习曲线: 平缓
-
-**中期** (3个月):
-- 技能数量: -55% (47 → 21)
-- 维护成本: -50%
-- 依赖关系: 清晰
-
-**长期** (6个月):
-- 技能体系: 稳定
-- 新技能集成: 标准化
-- 社区贡献: 增加
-
----
-
-## 下一步
-
-1. **Week 1**: 开始 Phase 1 核心合并
-2. **Week 2**: 完成功能合并
-3. **Week 3**: 完成集成合并
-4. **Week 4**: 验证和文档
-
----
-
-**创建**: 2026-04-01
-**状态**: 规划中
-**优先级**: P0
+---
+id: "cache-optimization-and-skill-integration"
+title: "缓存优化和技能整合 (2026-04-01)"
+type: "custom"
+tags: [memory, template]
+template: true
+---
+
+# 缓存优化和技能整合 (2026-04-01)
+
+## cc-cache-fix 集成
+
+**状态**: ✅ 已验证和集成
+
+**核心发现**:
+- cc-cache-fix 解决 60-70% token 燃烧问题（相对基准）
+- 三个补丁：Delta 附件持久化、Hash 稳定性、TTL 扩展
+- 需要配合 context 压缩、会话分割、缓存预热才能达到最优效果
+
+**立即应用**:
+```bash
+git clone https://github.com/Rangizingo/cc-cache-fix.git
+cd cc-cache-fix && ./install.sh
+alias claude=claude-patched
+```
+
+**文档**: `rules/domain/cache-management.md`
+
+---
+
+## 技能库整合计划
+
+**总体目标**: 47 个技能 → 21 个统一技能体系
+
+**分类**:
+- Agent 工程 (11) → 5 个合并技能
+- 搜索数据 (3) → 2 个合并技能
+- 创作设计 (4) → 3 个合并技能
+- 视频音频 (7) → 4 个合并技能
+- 浏览器爬虫 (3) → 1 个合并技能
+- 工具集成 (4) → 2 个合并技能
+- 安全运维 (3) → 1 个合并技能
+- 知识工具 (12) → 1 个独立仓库
+- 保持独立 (9) → 9 个独立技能
+
+**核心合并** (P0):
+1. context-optimization-suite (context-budget-analyzer + context-compressor)
+2. decision-making-framework (high-agency + structured-decision-alignment)
+3. memory-evolution-system (memory-manager + proactive-self-improving)
+4. quality-assurance-framework (adversarial-verification + self-rationalization-guard)
+5. cache-optimization-skill (cc-cache-fix 集成) ⭐ NEW
+
+**文档**: 
+- `skills/SKILL_INTEGRATION_MAP.md` - 技能整合地图
+- `skills/MIGRATION_GUIDE.md` - 迁移指南
+
+---
+
+## 工程化规则更新
+
+**已更新**:
+- `rules/domain/cache-management.md` - 新增缓存管理规则
+- `rules/performance.md` - 添加缓存优化优先级
+
+**需要更新**:
+- `rules/domain/engineering-workflows.md` - 添加缓存预热步骤
+- `CLAUDE.md` - 添加技能体系说明
+- `capabilities/skills-guide.md` - 更新技能列表
+
+---
+
+## 预期收益
+
+**短期** (1个月):
+- Token 节省: 60-70%
+- 技能数量: -30% (47 → 33)
+- 学习曲线: 平缓
+
+**中期** (3个月):
+- 技能数量: -55% (47 → 21)
+- 维护成本: -50%
+- 依赖关系: 清晰
+
+**长期** (6个月):
+- 技能体系: 稳定
+- 新技能集成: 标准化
+- 社区贡献: 增加
+
+---
+
+## 下一步
+
+1. **Week 1**: 开始 Phase 1 核心合并
+2. **Week 2**: 完成功能合并
+3. **Week 3**: 完成集成合并
+4. **Week 4**: 验证和文档
+
+---
+
+**创建**: 2026-04-01
+**状态**: 规划中
+**优先级**: P0

package/knowledge/memory/engineering-patterns.md

@@ -1,134 +1,134 @@
----
-id: "engineering-patterns"
-title: "Engineering Patterns Memory"
-type: "custom"
-tags: [memory, template]
-template: true
----
-
-# Engineering Patterns Memory
-> Consolidated from: engineering-workflows.md, coding.md, testing.md, security.md
-> Verified patterns only - each earned through real experience
-
----
-
-## Daily Workflow Patterns (Always Apply)
-
-### 1. UI Modification → Visual Verification Loop
-```
-Code change → npm run dev → Navigate → Verify render → Check console → Mark complete
-```
-**Never** mark UI tasks complete without visual verification.
-
-### 2. Feature Implementation → Dependency-Ordered Steps
-```
-Step 1: [task] — depends: none
-Step 2: [task] — depends: Step 1
-Step 3: [task] — depends: Step 1, 2
-```
-Break into steps, annotate dependencies, execute in order.
-
-### 3. Database Changes → Transactional
-```sql
-BEGIN;
--- changes
-SELECT ... -- verify
-COMMIT;  -- or ROLLBACK
-```
-Always through Bytebase MCP. Always include rollback.
-
----
-
-## Advanced Patterns (Trigger-Based)
-
-### TDD Full Flow (new features, critical logic)
-RED → GREEN → REFACTOR → Property tests → Mutation tests
-- Coverage ≥ 80%, all edge cases, critical paths have property tests
-
-### DB Migration Orchestration (schema changes)
-1. Dependency analysis → 2. Forward + rollback scripts → 3. Integrity check
-4. Backward-compatible views → 5. Test env first → 6. Prod with checkpoints
-
-### Self-Healing Components (core UI, design system)
-Scan → Regression tests → Responsive variants → Performance bench
-→ TypeScript interfaces → Storybook → Unit tests → Auto-fix
-
-### 3D Visualization (Three.js / R3F)
-- **Sector angles**: Weight by node count (50 Skills ≠ 3 Hooks)
-- **Overflow**: Multi-ring spiral when nodes > sector capacity
-- **Connection points**: Surface offset, NOT center
-  - Sphere: 1.0, Cube: 1.2, Torus: 1.4
-- **Verify**: Screenshot global layout + local connections
-
-### Big Data + Frontend
-- **Fetch**: Batch queries, controlled size
-- **Process**: Dedup (Set) + classify + aggregate
-- **Render**: Lazy load images, video `#t=0.1`, `object-fit: cover`
-- **Search**: Multi-field (username + numeric ID)
-- **Display names**: Always deduplicate for marquee/lists
-
----
-
-## Code Quality Standards
-
-### Immutability (CRITICAL)
-```typescript
-// ALWAYS new object, NEVER mutate
-return { ...user, name };  // NOT: user.name = name
-```
-
-### File Organization
-- 200-400 lines typical, 800 max
-- Organize by feature/domain, not by type
-- High cohesion, low coupling
-
-### Error Handling
-```typescript
-try {
-  return await riskyOperation();
-} catch (error) {
-  console.error('Failed:', error);
-  throw new Error('User-friendly message');
-}
-```
-
-### Input Validation (Zod at boundaries)
-```typescript
-const schema = z.object({
-  email: z.string().email(),
-  age: z.number().int().min(0).max(150),
-});
-```
-
-### Common Patterns
-- API Response: `{ success, data?, error?, meta? }`
-- Custom Hooks: useDebounce, useLocalStorage
-- Repository: `findAll, findById, create, update, delete`
-
----
-
-## Task Type → Workflow Matrix
-
-| Task | Minimum | Recommended |
-|------|---------|-------------|
-| Simple UI change | Visual verify | - |
-| New feature | Structured planning | + TDD |
-| DB change | Transactional | + Migration orchestration |
-| Core component | Visual verify + planning | + Self-healing |
-| Multi-file refactor | Structured planning | + TDD |
-| 3D visualization | Visual verify + 3D rules | Screenshot loop |
-| Big data display | Structured planning + data rules | Batch + dedup + lazy |
-
----
-
-## Security Checklist (Pre-Commit)
-- [ ] No hardcoded secrets (env vars only)
-- [ ] User inputs validated (Zod)
-- [ ] SQL injection prevention (parameterized)
-- [ ] XSS prevention (sanitized HTML)
-- [ ] Auth/authz verified
-- [ ] Rate limiting on endpoints
-- [ ] Error messages don't leak data
-
----
-_Updated: 2026-03-02 | Source: rules/domain/*.md consolidated_
+---
+id: "engineering-patterns"
+title: "Engineering Patterns Memory"
+type: "custom"
+tags: [memory, template]
+template: true
+---
+
+# Engineering Patterns Memory
+> Consolidated from: engineering-workflows.md, coding.md, testing.md, security.md
+> Verified patterns only - each earned through real experience
+
+---
+
+## Daily Workflow Patterns (Always Apply)
+
+### 1. UI Modification → Visual Verification Loop
+```
+Code change → npm run dev → Navigate → Verify render → Check console → Mark complete
+```
+**Never** mark UI tasks complete without visual verification.
+
+### 2. Feature Implementation → Dependency-Ordered Steps
+```
+Step 1: [task] — depends: none
+Step 2: [task] — depends: Step 1
+Step 3: [task] — depends: Step 1, 2
+```
+Break into steps, annotate dependencies, execute in order.
+
+### 3. Database Changes → Transactional
+```sql
+BEGIN;
+-- changes
+SELECT ... -- verify
+COMMIT;  -- or ROLLBACK
+```
+Always through Bytebase MCP. Always include rollback.
+
+---
+
+## Advanced Patterns (Trigger-Based)
+
+### TDD Full Flow (new features, critical logic)
+RED → GREEN → REFACTOR → Property tests → Mutation tests
+- Coverage ≥ 80%, all edge cases, critical paths have property tests
+
+### DB Migration Orchestration (schema changes)
+1. Dependency analysis → 2. Forward + rollback scripts → 3. Integrity check
+4. Backward-compatible views → 5. Test env first → 6. Prod with checkpoints
+
+### Self-Healing Components (core UI, design system)
+Scan → Regression tests → Responsive variants → Performance bench
+→ TypeScript interfaces → Storybook → Unit tests → Auto-fix
+
+### 3D Visualization (Three.js / R3F)
+- **Sector angles**: Weight by node count (50 Skills ≠ 3 Hooks)
+- **Overflow**: Multi-ring spiral when nodes > sector capacity
+- **Connection points**: Surface offset, NOT center
+  - Sphere: 1.0, Cube: 1.2, Torus: 1.4
+- **Verify**: Screenshot global layout + local connections
+
+### Big Data + Frontend
+- **Fetch**: Batch queries, controlled size
+- **Process**: Dedup (Set) + classify + aggregate
+- **Render**: Lazy load images, video `#t=0.1`, `object-fit: cover`
+- **Search**: Multi-field (username + numeric ID)
+- **Display names**: Always deduplicate for marquee/lists
+
+---
+
+## Code Quality Standards
+
+### Immutability (CRITICAL)
+```typescript
+// ALWAYS new object, NEVER mutate
+return { ...user, name };  // NOT: user.name = name
+```
+
+### File Organization
+- 200-400 lines typical, 800 max
+- Organize by feature/domain, not by type
+- High cohesion, low coupling
+
+### Error Handling
+```typescript
+try {
+  return await riskyOperation();
+} catch (error) {
+  console.error('Failed:', error);
+  throw new Error('User-friendly message');
+}
+```
+
+### Input Validation (Zod at boundaries)
+```typescript
+const schema = z.object({
+  email: z.string().email(),
+  age: z.number().int().min(0).max(150),
+});
+```
+
+### Common Patterns
+- API Response: `{ success, data?, error?, meta? }`
+- Custom Hooks: useDebounce, useLocalStorage
+- Repository: `findAll, findById, create, update, delete`
+
+---
+
+## Task Type → Workflow Matrix
+
+| Task | Minimum | Recommended |
+|------|---------|-------------|
+| Simple UI change | Visual verify | - |
+| New feature | Structured planning | + TDD |
+| DB change | Transactional | + Migration orchestration |
+| Core component | Visual verify + planning | + Self-healing |
+| Multi-file refactor | Structured planning | + TDD |
+| 3D visualization | Visual verify + 3D rules | Screenshot loop |
+| Big data display | Structured planning + data rules | Batch + dedup + lazy |
+
+---
+
+## Security Checklist (Pre-Commit)
+- [ ] No hardcoded secrets (env vars only)
+- [ ] User inputs validated (Zod)
+- [ ] SQL injection prevention (parameterized)
+- [ ] XSS prevention (sanitized HTML)
+- [ ] Auth/authz verified
+- [ ] Rate limiting on endpoints
+- [ ] Error messages don't leak data
+
+---
+_Updated: 2026-03-02 | Source: rules/domain/*.md consolidated_

package/knowledge/memory/feedback_root_structure.md

@@ -1,16 +1,16 @@
----
-id: "feedback_root_structure"
-title: "feedback_root_structure"
-type: "custom"
-tags: [memory, template]
-template: true
----
-
----
-name: follow-root-structure
-description: Enforce current root directory classification for new work
-type: feedback
----
-Always create/move new work to match the current root folder classification (00-Inbox, 01-Assets, 02-Docs, 03-Scripts, 04-Configs, 05-Outputs, 06-Archive, 99-Projects-Other, bo-work, bo-skill-research, etc.).
-**Why:** User wants strict adherence to the newly organized root structure for all future Claude Code actions.
+---
+id: "feedback_root_structure"
+title: "feedback_root_structure"
+type: "custom"
+tags: [memory, template]
+template: true
+---
+
+---
+name: follow-root-structure
+description: Enforce current root directory classification for new work
+type: feedback
+---
+Always create/move new work to match the current root folder classification (00-Inbox, 01-Assets, 02-Docs, 03-Scripts, 04-Configs, 05-Outputs, 06-Archive, 99-Projects-Other, bo-work, bo-skill-research, etc.).
+**Why:** User wants strict adherence to the newly organized root structure for all future Claude Code actions.
 **How to apply:** When creating files/projects or moving outputs, place them into the appropriate existing bucket instead of adding new top-level folders or leaving items at root.