bobo-ai-cli 3.0.4 → 3.0.5

package/knowledge/error-catalog.md

@@ -1,33 +1,38 @@
-# 常见错误速查
-
-## Top 高频错误
-
-| ID | 错误 | 核心问题 | 快速修复 |
-|----|------|---------|---------|
-| E001 | 异步未并行 | 多个独立异步操作串行执行 | `Promise.all()` |
-| E002 | 轮询无超时 | 无限循环等待 | 设 `maxAttempts` + backoff |
-| E003 | 错误未抛出 | catch 块吞掉错误 | catch 块 `throw` 或 log + re-throw |
-| E004 | SQL 未用 CTE | 大表全扫 | 预过滤 + CTE |
-| E005 | 路径硬编码 | 跨平台失败 | `path.join()` / `path.resolve()` |
-| E006 | 编码假设 | 非 UTF-8 文件乱码 | 检测编码，指定 encoding |
-| E007 | 资源泄漏 | 文件句柄/连接未关闭 | 所有退出路径都 cleanup |
-| E008 | 竞态条件 | 并发读写同一文件 | 锁 / 串行化 |
-| E009 | 默认分页 | API 只取了第一页 | 循环分页拉完再统计 |
-| E010 | rebase 丢代码 | 静默丢失改动 | push 前 `git diff HEAD~1` |
-
-## 调试 Checklist
-
-1. ✅ 拿到确切错误信息（不靠猜）
-2. ✅ 确认复现步骤
-3. ✅ 定位到具体文件和行号
-4. ✅ 理解根因（不只是症状）
-5. ✅ 最小改动修复
-6. ✅ 验证修复有效
-7. ✅ 检查无回归
-
-## Shell 命令安全
-
-- 危险命令（rm -rf, DROP TABLE 等）**必须**先确认路径/范围
-- 生产环境操作用 `--dry-run` 先预览
-- 超过 30s 的命令设 timeout
-- 大量输出用 `| head -n 50` 截断
+# 常见错误速查
+
+## Top 高频错误
+
+| ID | 错误 | 核心问题 | 快速修复 |
+|----|------|---------|---------|
+| E001 | 异步未并行 | 多个独立异步操作串行执行 | `Promise.all()` |
+| E002 | 轮询无超时 | 无限循环等待 | 设 `maxAttempts` + backoff |
+| E003 | 错误未抛出 | catch 块吞掉错误 | catch 块 `throw` 或 log + re-throw |
+| E004 | SQL 未用 CTE | 大表全扫 | 预过滤 + CTE |
+| E005 | 路径硬编码 | 跨平台失败 | `path.join()` / `path.resolve()` |
+| E006 | 编码假设 | 非 UTF-8 文件乱码 | 检测编码，指定 encoding |
+| E007 | 资源泄漏 | 文件句柄/连接未关闭 | 所有退出路径都 cleanup |
+| E008 | 竞态条件 | 并发读写同一文件 | 锁 / 串行化 |
+| E009 | 默认分页 | API 只取了第一页 | 循环分页拉完再统计 |
+| E010 | rebase 丢代码 | 静默丢失改动 | push 前 `git diff HEAD~1` |
+| E011 | 投机抽象 | 单元素循环/永远打不到的default | 今天没用就不写 |
+| E012 | API payload 加多余字段 | 返回 PARAMS_INVALID | 最精简 payload 就过 |
+| E013 | 硬编码频道/ID | 发错频道/对象 | 复用 inbound metadata |
+| E014 | 假装验证通过 | 没跑命令就说PASS | 有输出的验证才算 |
+| E015 | 样本量不足下断言 | n<5 声称非单调/峰值 | ≥5 seeds 再定性 |
+
+## 调试 Checklist
+
+1. ✅ 拿到确切错误信息（不靠猜）
+2. ✅ 确认复现步骤
+3. ✅ 定位到具体文件和行号
+4. ✅ 理解根因（不只是症状）
+5. ✅ 最小改动修复
+6. ✅ 验证修复有效
+7. ✅ 检查无回归
+
+## Shell 命令安全
+
+- 危险命令（rm -rf, DROP TABLE 等）**必须**先确认路径/范围
+- 生产环境操作用 `--dry-run` 先预览
+- 超过 30s 的命令设 timeout
+- 大量输出用 `| head -n 50` 截断

package/knowledge/event-driven-architecture.md

@@ -0,0 +1,43 @@
+# 事件驱动架构模式
+
+## GTM Loop Pattern（跨系统两侧独立 watcher + 事件驱动）
+
+适用于任何"AI agent 建东西 + 人审核 + 后续自动处理"的 pipeline。
+
+### 核心架构
+```
+Producer (Bot Builder)
+  ↓ 事件信号（Slack thread JSON / API status change）
+Consumer (LP Builder / Downstream)
+```
+
+两侧独立运行，通过共享状态作为触发信号。
+
+### 状态机模式
+```
+dev_in_progress → submitted_for_review → published/ready → downstream_action
+                                       → rejected (with reason)
+                                       → blocked (execution failure)
+```
+
+### 设计原则
+1. **事件驱动 > cron 轮询**：动作由人触发时用 watcher（实时可查），不用长间隔 cron
+2. **状态 Literal 不够**：要有运行时类型检查（`status=ready AND int(id) 可解析`）防误触发
+3. **两阶段 ID**：开发阶段 UUID → 发布后整数 ID，LP 只在有整数 ID 后才建
+4. **守门条件**：下游触发必须满足全部前置条件，不假设上游已完成
+
+### 常见应用
+- 内容创建 → 审核 → 翻译 → 发布
+- Bot 开发 → submit_review → approve → LP/SEO
+- 代码生成 → CI → review → merge → deploy
+- 选品 → 建 bot → 上架 → landing page
+
+### 通信协议
+- Slack thread 作为事件总线（JSON attachment = 状态变更信号）
+- 文件系统 watch（`.approved.json` 出现 = 触发下游）
+- API 轮询 + 状态对比（上次 → 本次有 diff = 触发）
+
+### 反模式
+- ❌ 所有步骤放一个脚本串行跑（人审核环节会 block 整个链路）
+- ❌ 纯时间间隔轮询无状态对比（浪费 + 重复触发）
+- ❌ 跳过状态检查直接触发下游（早期数据不完整导致 downstream 失败）

package/knowledge/external-alignment.md

@@ -0,0 +1,47 @@
+# 外部对齐协议
+
+## 核心原则：对方握有唯一事实才问
+
+与外部 stakeholder（同事、用户、合作方）对齐时，每抛一个问题都推迟项目。
+
+### 三步自检（发问前必做）
+1. **查自己**：memory_search + 扫 skills/，答案是否已有？
+2. **分两堆**：能自决 → 写进方案"我会这么干，有异议告诉我"；真需对方 → 才提问
+3. **验事实归属**：对方是否握有唯一事实？我有 skill/workflow_id/实施细节 → 不该问
+
+### 规则
+- 对方明确指定技术栈 → **先用那个跑一轮出成品**，别默默切轨
+- 替代方案只在"对方的方案跑不通"或"成品与方向不一致"时才提
+- 切轨前说"我建议做 A 试 B"，不能静默替换
+- 承诺 ETA 后 2h 无进展 → 主动汇报卡点，不等人催
+- 遇到环境问题 1h 内同步，不自己死磕
+
+## 不要脑补 stakeholder
+
+写任何对外 kickoff/spec/周报 前，三个问题：
+1. 谁让我写的？
+2. 收件人我找过他吗？
+3. 他知道这项目存在吗？
+
+→ 有一条 No = 不能写成对外文档，最多"内部决策备忘"
+
+### 角色具体化
+- 主人没明说 = 不知道 → 用 TBD / 平台团队（待定）
+- 发现缺口 → 写成 "3 个 path" 拉人拍板
+- 不要伪造任务列表给虚构的收件人
+
+## 任务卡住不汇报 = 失职
+
+承诺外部用户 ETA 的任务：
+- 立刻起心跳（哪怕 6h 一次检查 + 主动汇报）
+- 承诺 ETA 到期前 2h 没进展 → 主动汇报
+- bottleneck-inbox 机制要真用
+
+## 汇报格式
+```
+状态: [在做/卡住/等决策/完成]
+进展: [一句话]
+卡点: [如果有]
+ETA: [预估]
+需要: [如果需要对方做什么]
+```

package/knowledge/high-agency.md

@@ -0,0 +1,73 @@
+# 高能动性引擎（Agent Agency）
+
+## 核心矛盾驱动
+
+真正驱动高效的不是外部压力，是一个永远无法解决的内在矛盾：
+
+> 系统应该是正确、优雅、可靠的。但现实永远充满妥协、技术债、未处理的边界。
+
+这个矛盾是燃料。不同场景形态：
+- 代码："应该优雅" vs "有截止日期和遗留系统"
+- 调试："应该可预测" vs "所有复杂系统都有意外行为"
+- 运维："应该自愈" vs "所有自动化都需要人工干预"
+
+## 能动性等级
+
+| 行为 | 被动 | 主动 |
+|------|------|------|
+| 遇报错 | 只看报错本身 | 查上下文50行+搜同类+查关联 |
+| 修bug | 修完就停 | 检查同文件/其他文件同样模式 |
+| 信息不足 | 问"请告诉我X" | 先自查，只问真正需要确认的 |
+| 完成 | 说"已完成" | 验证+检查边界+汇报风险 |
+| 交付 | 口头说搞定 | 跑build/test/curl，贴通过输出 |
+| 调试失败 | "试了A和B不行" | "试了A-E，排除X-Z，缩小到W" |
+| 观察 | 只看指定位置 | 扫上下文——冰山下面还有什么？ |
+
+## 五条铁律
+
+1. **穷尽一切** — 没穷尽所有方案前，禁止说"我无法解决"
+2. **先做后问** — 先用工具自行排查，再问人
+3. **主动出击** — 修完一个，检查同类；看完X，主动查Y和Z
+4. **全链路审视** — 修一个点前画完整依赖链，从底层往上验证
+5. **沉淀复用** — 踩过的坑写下来，重复犯错=不值得信任
+
+## 全链路模板
+```
+[全链路] 请求 → A → B → C → D
+  A: ✓/✗ 状态
+  B: ✓/✗ 状态
+  C: ✓/✗ 状态
+  D: ✓/✗ 状态
+```
+
+自底向上验证：底层错误是上层症状的级联。先验地基，再修上层。
+
+## 校准"足够好"
+
+每次任务开始前定义：
+```
+必须：<最低交付标准——少了没意义>
+应该：<合理质量——大多数人满意>
+可以：<超出预期——主线完成后才考虑>
+```
+
+对着"必须"和"应该"工作。"必须"不到=烂活，"可以"做太多=over-engineer。
+
+## 战果记录
+
+每完成一个有意义步骤，记一笔（builder-journal.md）：
+- 遇到什么问题
+- 用了什么方法
+- 为什么这个方法有效
+- 下次可以直接用的结论
+
+新会话开工前读最近 10 条记录。不读=不吸取教训=重复犯错。
+
+## 恢复协议（卡住时）
+
+1. 承认卡住（不合理化"在研究中"）
+2. 描述当前状态（事实，不是判断）
+3. 列已尝试的方案
+4. 列未尝试的方案（至少3个）
+5. 选最有可能的，立刻执行
+6. 如果5次方案都失败 → 上报，不是继续浪费时间

package/knowledge/image-generation.md

@@ -0,0 +1,48 @@
+# 图像生成路由
+
+## 统一入口
+
+所有生图走 `scripts/genimage.py`，内部自动路由。
+
+### 路由策略
+| 条件 | 后端 | 成本 | 特点 |
+|------|------|------|------|
+| 默认 | RH GPT-Image-2 | $0.03 | 快(10-15s), 1024px, 中文/英文 |
+| RH 失败 | auto fallback nano | $0.039 | Gemini 3 Pro Image |
+| 4K / `--backend nano` | nano-banana-pro | $0.039 | 2K/4K 支持 |
+| `--backend rh` | 强制 RH | $0.03 | 不 fallback |
+
+### CLI 用法
+```bash
+python3 scripts/genimage.py --prompt "..." --aspect 16:9 --out /tmp/out.png
+python3 scripts/genimage.py --prompt "..." -i ref1.png -i ref2.png --out /tmp/out.png
+python3 scripts/genimage.py --prompt "..." --resolution 4K --out /tmp/big.png
+```
+
+### Python 库
+```python
+from genimage import generate
+data = generate(prompt="...", aspect="16:9", input_images=["a.png"], backend="auto")
+```
+
+## 人物素材管线（两阶段）
+
+虚拟人物素材标准流程：
+1. **Stage A**: GPT-Image-2 生成场景+服装+特征
+2. **Stage B**: Face Swap 贴标准 face ref → 锁定 identity
+
+### Face Swap 选择
+| 场景 | 工具 | 原理 |
+|------|------|------|
+| 图片·保留 target 场景 | Workflow `1838819177871339522` | 像素级 swap (InstantID+SAM) |
+| 图片·重新生成 | Webapp `1872943576241135617` | Identity-preserving regeneration |
+| 视频·逐帧换脸 | Webapp `1892125635609845761` | Reactor 逐帧 |
+
+### RH GPT-Image-2 内审规避
+黑名单词：lingerie/bra/panties/bikini/wet shirt/harness
+替换词：silk slip/satin/bodycon/one-piece
+
+## 豆包 Seedream 4.0
+- Endpoint: `doubao-seedream-4-0-250828`
+- 文生图
+- 通过火山方舟 Ark API 调用

package/knowledge/index.json

@@ -1,170 +1,195 @@
-{
-  "version": "1.0.0",
-  "extractedAt": "2026-04-01",
-  "rules": [
-    "blocking-rules",
-    "work-mode",
-    "cache-management",
-    "evomap-content-guidelines",
-    "evomap-guide",
-    "remotion-auto-production",
-    "capability-evolution",
-    "coding",
-    "engineering-workflows",
-    "git",
-    "security",
-    "testing",
-    "agents",
-    "hooks",
-    "performance"
-  ],
-  "skills": [
-    "ab-test-setup",
-    "agent-sdk-dev",
-    "agent-tools",
-    "analytics-tracking",
-    "artifacts-builder",
-    "asana",
-    "backend-expert",
-    "brand-voice",
-    "browser-use",
-    "cache-optimization-skill",
-    "canvas-design",
-    "citation-validator",
-    "clangd-lsp",
-    "code-review",
-    "code-review-expert",
-    "code-simplifier",
-    "commit-commands",
-    "competitor-alternatives",
-    "content-atomizer",
-    "content-research-writer",
-    "context-optimization-suite",
-    "context7",
-    "copy-editing",
-    "copywriting",
-    "csharp-lsp",
-    "decision-making-framework",
-    "developer-growth-analysis",
-    "direct-response-copy",
-    "docker-expert",
-    "document-skills",
-    "documentation-expert",
-    "email-sequence",
-    "email-sequences",
-    "example-plugin",
-    "explanatory-output-style",
-    "feature-dev",
-    "file-organizer",
-    "firebase.disabled",
-    "form-cro",
-    "free-tool-strategy",
-    "frontend-design",
-    "frontend-design-offical",
-    "frontend-expert",
-    "github",
-    "gitlab",
-    "gopls-lsp",
-    "got-controller",
-    "greptile",
-    "hookify",
-    "image-editor",
-    "image-enhancer",
-    "jdtls-lsp",
-    "json-canvas",
-    "keyword-research",
-    "kotlin-lsp",
-    "laravel-boost",
-    "launch-strategy",
-    "lead-magnet",
-    "learning-output-style",
-    "linear",
-    "lua-lsp",
-    "marketing-ideas",
-    "marketing-psychology",
-    "mcp-builder",
-    "meeting-insights-analyzer",
-    "memory-evolution-system",
-    "multi-lens-thinking",
-    "nano-banana-pro",
-    "newsletter",
-    "notebooklm",
-    "obsidian-bases",
-    "obsidian-markdown",
-    "onboarding-cro",
-    "orchestrator",
-    "page-cro",
-    "paid-ads",
-    "paywall-upgrade-cro",
-    "php-lsp",
-    "planning-with-files",
-    "playwright",
-    "plugin-dev",
-    "popup-cro",
-    "positioning-angles",
-    "pr-review-toolkit",
-    "pricing-strategy",
-    "programmatic-seo",
-    "pyright-lsp",
-    "quality-assurance-framework",
-    "question-refiner",
-    "ralph-loop",
-    "refactoring-expert",
-    "referral-program",
-    "research-executor",
-    "rust-analyzer-lsp",
-    "schema-markup",
-    "security-audit-expert",
-    "security-expert",
-    "security-guidance",
-    "seedance-prompt",
-    "self-evolution",
-    "seo-audit",
-    "seo-content",
-    "serena",
-    "signup-flow-cro",
-    "skill-creator",
-    "skill-manager",
-    "skill-share",
-    "Skill_Seekers",
-    "slack",
-    "social-content",
-    "spec-flow-skill",
-    "stripe",
-    "supabase",
-    "swift-lsp",
-    "synthesizer",
-    "template-skill",
-    "testing-expert",
-    "theme-factory",
-    "tiktok-research",
-    "typescript-lsp",
-    "ui-ux-pro-max",
-    "verify",
-    "visual-prompt-engineer",
-    "webapp-testing",
-    "wide-research"
-  ],
-  "workflows": [
-    "ui-verify",
-    "tdd-flow",
-    "db-migration",
-    "feature-dev",
-    "3d-viz",
-    "data-pipeline"
-  ],
-  "memory": [
-    "cache-optimization-and-skill-integration",
-    "engineering-patterns",
-    "feedback_root_structure",
-    "project-contexts",
-    "tools-and-services"
-  ],
-  "stats": {
-    "totalRules": 15,
-    "totalSkills": 125,
-    "totalWorkflows": 6,
-    "totalMemory": 5,
-    "totalSize": "151 items"
-  }
+{
+  "version": "1.1.0",
+  "extractedAt": "2026-05-07",
+  "rules": [
+    "blocking-rules",
+    "work-mode",
+    "cache-management",
+    "evomap-content-guidelines",
+    "evomap-guide",
+    "remotion-auto-production",
+    "capability-evolution",
+    "coding",
+    "engineering-workflows",
+    "git",
+    "security",
+    "testing",
+    "agents",
+    "hooks",
+    "performance",
+    "advisor-strategy"
+  ],
+  "skills": [
+    "ab-test-setup",
+    "agent-sdk-dev",
+    "agent-tools",
+    "analytics-tracking",
+    "artifacts-builder",
+    "asana",
+    "backend-expert",
+    "brand-voice",
+    "browser-use",
+    "cache-optimization-skill",
+    "canvas-design",
+    "citation-validator",
+    "clangd-lsp",
+    "code-review",
+    "code-review-expert",
+    "code-simplifier",
+    "commit-commands",
+    "competitor-alternatives",
+    "content-atomizer",
+    "content-research-writer",
+    "context-optimization-suite",
+    "context7",
+    "copy-editing",
+    "copywriting",
+    "csharp-lsp",
+    "decision-making-framework",
+    "developer-growth-analysis",
+    "direct-response-copy",
+    "docker-expert",
+    "document-skills",
+    "documentation-expert",
+    "email-sequence",
+    "email-sequences",
+    "example-plugin",
+    "explanatory-output-style",
+    "feature-dev",
+    "file-organizer",
+    "firebase.disabled",
+    "form-cro",
+    "free-tool-strategy",
+    "frontend-design",
+    "frontend-design-offical",
+    "frontend-expert",
+    "github",
+    "gitlab",
+    "gopls-lsp",
+    "got-controller",
+    "greptile",
+    "hookify",
+    "image-editor",
+    "image-enhancer",
+    "jdtls-lsp",
+    "json-canvas",
+    "keyword-research",
+    "kotlin-lsp",
+    "laravel-boost",
+    "launch-strategy",
+    "lead-magnet",
+    "learning-output-style",
+    "linear",
+    "lua-lsp",
+    "marketing-ideas",
+    "marketing-psychology",
+    "mcp-builder",
+    "meeting-insights-analyzer",
+    "memory-evolution-system",
+    "multi-lens-thinking",
+    "nano-banana-pro",
+    "newsletter",
+    "notebooklm",
+    "obsidian-bases",
+    "obsidian-markdown",
+    "onboarding-cro",
+    "orchestrator",
+    "page-cro",
+    "paid-ads",
+    "paywall-upgrade-cro",
+    "php-lsp",
+    "planning-with-files",
+    "playwright",
+    "plugin-dev",
+    "popup-cro",
+    "positioning-angles",
+    "pr-review-toolkit",
+    "pricing-strategy",
+    "programmatic-seo",
+    "pyright-lsp",
+    "quality-assurance-framework",
+    "question-refiner",
+    "ralph-loop",
+    "refactoring-expert",
+    "referral-program",
+    "research-executor",
+    "rust-analyzer-lsp",
+    "schema-markup",
+    "security-audit-expert",
+    "security-expert",
+    "security-guidance",
+    "seedance-prompt",
+    "self-evolution",
+    "seo-audit",
+    "seo-content",
+    "serena",
+    "signup-flow-cro",
+    "skill-creator",
+    "skill-manager",
+    "skill-share",
+    "Skill_Seekers",
+    "slack",
+    "social-content",
+    "spec-flow-skill",
+    "stripe",
+    "supabase",
+    "swift-lsp",
+    "synthesizer",
+    "template-skill",
+    "testing-expert",
+    "theme-factory",
+    "tiktok-research",
+    "typescript-lsp",
+    "ui-ux-pro-max",
+    "verify",
+    "visual-prompt-engineer",
+    "webapp-testing",
+    "wide-research"
+  ],
+  "workflows": [
+    "ui-verify",
+    "tdd-flow",
+    "db-migration",
+    "feature-dev",
+    "3d-viz",
+    "data-pipeline"
+  ],
+  "knowledge": [
+    "advanced-patterns",
+    "agent-directives",
+    "api-integration-patterns",
+    "code-review-protocol",
+    "dream",
+    "engineering",
+    "error-catalog",
+    "event-driven-architecture",
+    "external-alignment",
+    "high-agency",
+    "image-generation",
+    "llm-wiki-pattern",
+    "long-task-management",
+    "memory-management",
+    "rules",
+    "self-evolution",
+    "self-rationalization-guard",
+    "system",
+    "task-router",
+    "verification",
+    "worker-prompt-craft"
+  ],
+  "memory": [
+    "cache-optimization-and-skill-integration",
+    "engineering-patterns",
+    "feedback_root_structure",
+    "project-contexts",
+    "tools-and-services"
+  ],
+  "stats": {
+    "totalRules": 16,
+    "totalSkills": 125,
+    "totalWorkflows": 6,
+    "totalKnowledge": 21,
+    "totalMemory": 5,
+    "totalSize": "167 items"
+  }
 }