# liteagents-skill-installer - Doramagic AI Context Pack

> 定位：安装前体验与判断资产。它帮助宿主 AI 有一个好的开始，但不代表已经安装、执行或验证目标项目。

## 充分原则

- **充分原则，不是压缩原则**：AI Context Pack 应该充分到让宿主 AI 在开工前理解项目价值、能力边界、使用入口、风险和证据来源；它可以分层组织，但不以最短摘要为目标。
- **压缩策略**：只压缩噪声和重复内容，不压缩会影响判断和开工质量的上下文。

## 给宿主 AI 的使用方式

你正在读取 Doramagic 为 liteagents-skill-installer 编译的 AI Context Pack。请把它当作开工前上下文：帮助用户理解适合谁、能做什么、如何开始、哪些必须安装后验证、风险在哪里。不要声称你已经安装、运行或执行了目标项目。

## Claim 消费规则

- **事实来源**：Repo Evidence + Claim/Evidence Graph；Human Wiki 只提供显著性、术语和叙事结构。
- **事实最低状态**：`supported`
- `supported`：可以作为项目事实使用，但回答中必须引用 claim_id 和证据路径。
- `weak`：只能作为低置信度线索，必须要求用户继续核实。
- `inferred`：只能用于风险提示或待确认问题，不能包装成项目事实。
- `unverified`：不得作为事实使用，应明确说证据不足。
- `contradicted`：必须展示冲突来源，不得替用户强行选择一个版本。

## 它最适合谁

- **正在使用 Claude/Codex/Cursor/Gemini 等宿主 AI 的开发者**：README 或插件配置提到多个宿主 AI。 证据：`README.md` Claim：`clm_0004` supported 0.86
- **希望把专业流程带进宿主 AI 的用户**：仓库包含 Skill 文档。 证据：`packages/claude/skills/brainstorming/SKILL.md`, `packages/claude/skills/debug-method/SKILL.md`, `packages/claude/skills/docs-builder/SKILL.md`, `packages/claude/skills/live-canvas/SKILL.md` 等 Claim：`clm_0005` supported 0.86

## 它能做什么

- **AI Skill / Agent 指令资产库**（可做安装前预览）：项目包含可被宿主 AI 读取的 Skill 或 Agent 指令文件，可用于把专业流程带入 Claude、Codex、Cursor 等宿主。 证据：`packages/claude/skills/brainstorming/SKILL.md`, `packages/claude/skills/debug-method/SKILL.md`, `packages/claude/skills/docs-builder/SKILL.md`, `packages/claude/skills/live-canvas/SKILL.md` 等 Claim：`clm_0001` supported 0.86
- **多宿主安装与分发**（需要安装后验证）：项目包含插件或 marketplace 配置，说明它面向一个或多个 AI 宿主的安装和分发。 证据：`packages/claude/plugins/live-canvas-marketplace/.claude-plugin/marketplace.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/.claude-plugin/plugin.json` Claim：`clm_0002` supported 0.86
- **命令行启动或安装流程**（需要安装后验证）：项目文档中存在可执行命令，真实使用需要在本地或宿主环境中运行这些命令。 证据：`README.md`, `packages/ampcode/commands/live-canvas/README.md`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md` Claim：`clm_0003` supported 0.86

## 怎么开始

- `npx liteagents` 证据：`README.md` Claim：`clm_0006` supported 0.86
- `npm install -g liteagents` 证据：`README.md` Claim：`clm_0007` supported 0.86
- `/plugin marketplace add ~/.claude/plugins/live-canvas-marketplace` 证据：`packages/ampcode/commands/live-canvas/README.md` Claim：`clm_0008` supported 0.86
- `/plugin install live-canvas-channel@live-canvas-marketplace` 证据：`packages/ampcode/commands/live-canvas/README.md` Claim：`clm_0009` supported 0.86
- `curl http://localhost:8788/health` 证据：`packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md` Claim：`clm_0010` supported 0.86
- `curl -X POST http://localhost:8788/feedback \` 证据：`packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md` Claim：`clm_0011` supported 0.86

## 继续前判断卡

- **当前建议**：先做流程沙盒试用
- **为什么**：这个项目会改变宿主 AI 的开发流程和规则，值得继续试，但不要直接装进主力 Claude/Cursor/Codex；先用临时宿主或隔离目录验证。

### 30 秒判断

- **现在怎么做**：先做流程沙盒试用
- **最小安全下一步**：先用 Prompt Preview 感受流程约束；满意后再临时宿主试装
- **先别相信**：这套流程是否适合你的工作方式不能直接相信。
- **继续会触碰**：宿主行为改变、命令执行、宿主 AI 配置

### 现在可以相信

- **适合人群线索：正在使用 Claude/Codex/Cursor/Gemini 等宿主 AI 的开发者**（supported）：有 supported claim 或项目证据支撑，但仍不等于真实安装效果。 证据：`README.md` Claim：`clm_0004` supported 0.86
- **适合人群线索：希望把专业流程带进宿主 AI 的用户**（supported）：有 supported claim 或项目证据支撑，但仍不等于真实安装效果。 证据：`packages/claude/skills/brainstorming/SKILL.md`, `packages/claude/skills/debug-method/SKILL.md`, `packages/claude/skills/docs-builder/SKILL.md`, `packages/claude/skills/live-canvas/SKILL.md` 等 Claim：`clm_0005` supported 0.86
- **能力存在：AI Skill / Agent 指令资产库**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`packages/claude/skills/brainstorming/SKILL.md`, `packages/claude/skills/debug-method/SKILL.md`, `packages/claude/skills/docs-builder/SKILL.md`, `packages/claude/skills/live-canvas/SKILL.md` 等 Claim：`clm_0001` supported 0.86
- **能力存在：多宿主安装与分发**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`packages/claude/plugins/live-canvas-marketplace/.claude-plugin/marketplace.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/.claude-plugin/plugin.json` Claim：`clm_0002` supported 0.86
- **能力存在：命令行启动或安装流程**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`README.md`, `packages/ampcode/commands/live-canvas/README.md`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md` Claim：`clm_0003` supported 0.86
- **存在 Quick Start / 安装命令线索**（supported）：可以相信项目文档出现过启动或安装入口；不要因此直接在主力环境运行。 证据：`README.md` Claim：`clm_0006` supported 0.86

### 现在还不能相信

- **这套流程是否适合你的工作方式不能直接相信。**（unverified）：流程型 Skill 可能强约束 AI 行为；它能提升纪律，也可能拖慢你当前任务节奏。 证据：`CLAUDE.md`, `packages/claude/plugins/live-canvas-marketplace/.claude-plugin/marketplace.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/.claude-plugin/plugin.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md` 等
- **不会和你已有 Claude/Cursor/Codex 规则冲突，不能直接相信。**（inferred）：开发流程 Skill 会改变澄清、计划、测试、验证等默认行为，必须在临时宿主里试。 证据：`CLAUDE.md`, `packages/claude/plugins/live-canvas-marketplace/.claude-plugin/marketplace.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/.claude-plugin/plugin.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md` 等
- **真实输出质量不能在安装前相信。**（unverified）：Prompt Preview 只能展示引导方式，不能证明真实项目中的结果质量。
- **宿主 AI 版本兼容性不能在安装前相信。**（unverified）：Claude、Cursor、Codex、Gemini 等宿主加载规则和版本差异必须在真实环境验证。
- **不会污染现有宿主 AI 行为，不能直接相信。**（inferred）：Skill、plugin、AGENTS/CLAUDE/GEMINI 指令可能改变宿主 AI 的默认行为。 证据：`CLAUDE.md`, `packages/claude/plugins/live-canvas-marketplace/.claude-plugin/marketplace.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/.claude-plugin/plugin.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md` 等
- **可安全回滚不能默认相信。**（unverified）：除非项目明确提供卸载和恢复说明，否则必须先在隔离环境验证。
- **真实安装后是否与用户当前宿主 AI 版本兼容？**（unverified）：兼容性只能通过实际宿主环境验证。 证据：`packages/claude/plugins/live-canvas-marketplace/.claude-plugin/marketplace.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/.claude-plugin/plugin.json`
- **项目输出质量是否满足用户具体任务？**（unverified）：安装前预览只能展示流程和边界，不能替代真实评测。

### 继续会触碰什么

- **宿主行为改变**：澄清、计划、TDD、验证、收尾等默认开发节奏。 原因：这类 Skill 的价值和风险都来自强约束流程；必须先确认你愿意被它改变工作方式。 证据：`CLAUDE.md`, `packages/claude/plugins/live-canvas-marketplace/.claude-plugin/marketplace.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/.claude-plugin/plugin.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md` 等
- **命令执行**：包管理器、网络下载、本地插件目录、项目配置或用户主目录。 原因：运行第一条命令就可能产生环境改动；必须先判断是否值得跑。 证据：`README.md`, `packages/ampcode/commands/live-canvas/README.md`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md`
- **宿主 AI 配置**：Claude/Codex/Cursor/Gemini/OpenCode 等宿主的 plugin、Skill 或规则加载配置。 原因：宿主配置会改变 AI 后续工作方式，可能和用户已有规则冲突。 证据：`CLAUDE.md`, `packages/claude/plugins/live-canvas-marketplace/.claude-plugin/marketplace.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/.claude-plugin/plugin.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md` 等
- **本地环境或项目文件**：安装结果、插件缓存、项目配置或本地依赖目录。 原因：安装前无法证明写入范围和回滚方式，需要隔离验证。 证据：`README.md`, `packages/ampcode/commands/live-canvas/README.md`, `packages/claude/plugins/live-canvas-marketplace/.claude-plugin/marketplace.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/.claude-plugin/plugin.json` 等
- **宿主 AI 上下文**：AI Context Pack、Prompt Preview、Skill 路由、风险规则和项目事实。 原因：导入上下文会影响宿主 AI 后续判断，必须避免把未验证项包装成事实。

### 最小安全下一步

- **先跑 Prompt Preview**：先感受它会怎样改变 AI 的开发节奏，再决定是否让它进入真实宿主。（适用：任何项目都适用，尤其是输出质量未知时。）
- **只在隔离目录或测试账号试装**：避免安装命令污染主力宿主 AI、真实项目或用户主目录。（适用：存在命令执行、插件配置或本地写入线索时。）
- **先备份宿主 AI 配置**：Skill、plugin、规则文件可能改变 Claude/Cursor/Codex 的默认行为。（适用：存在插件 manifest、Skill 或宿主规则入口时。）
- **安装后只验证一个最小任务**：先验证加载、兼容、输出质量和回滚，再决定是否深用。（适用：准备从试用进入真实工作流时。）

### 退出方式

- **保留安装前状态**：记录原始宿主配置和项目状态，后续才能判断是否可恢复。
- **准备移除宿主 plugin / Skill / 规则入口**：如果试装后行为异常，可以把宿主 AI 恢复到试装前状态。
- **记录安装命令和写入路径**：没有明确卸载说明时，至少要知道哪些目录或配置需要手动清理。
- **如果没有回滚路径，不进入主力环境**：不可回滚是继续前阻断项，不应靠信任或运气继续。

## 哪些只能预览

- 解释项目适合谁和能做什么
- 基于项目文档演示典型对话流程
- 帮助用户判断是否值得安装或继续研究

## 哪些必须安装后验证

- 真实安装 Skill、插件或 CLI
- 执行脚本、修改本地文件或访问外部服务
- 验证真实输出质量、性能和兼容性

## 边界与风险判断卡

- **把安装前预览误认为真实运行**：用户可能高估项目已经完成的配置、权限和兼容性验证。 处理方式：明确区分 prompt_preview_can_do 与 runtime_required。 Claim：`clm_0012` inferred 0.45
- **宿主 AI 插件或 Skill 规则冲突**：新规则可能改变用户现有宿主 AI 的工作方式。 处理方式：安装前先检查插件 manifest 和 Skill 文件，必要时隔离测试。 证据：`packages/claude/plugins/live-canvas-marketplace/.claude-plugin/marketplace.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/.claude-plugin/plugin.json` Claim：`clm_0013` supported 0.86
- **命令执行会修改本地环境**：安装命令可能写入用户主目录、宿主插件目录或项目配置。 处理方式：先在隔离环境或测试账号中运行。 证据：`README.md`, `packages/ampcode/commands/live-canvas/README.md`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md` Claim：`clm_0014` supported 0.86
- **待确认**：真实安装后是否与用户当前宿主 AI 版本兼容？。原因：兼容性只能通过实际宿主环境验证。
- **待确认**：项目输出质量是否满足用户具体任务？。原因：安装前预览只能展示流程和边界，不能替代真实评测。
- **待确认**：安装命令是否需要网络、权限或全局写入？。原因：这影响企业环境和个人环境的安装风险。

## 开工前工作上下文

### 加载顺序

- 先读取 how_to_use.host_ai_instruction，建立安装前判断资产的边界。
- 读取 claim_graph_summary，确认事实来自 Claim/Evidence Graph，而不是 Human Wiki 叙事。
- 再读取 intended_users、capabilities 和 quick_start_candidates，判断用户是否匹配。
- 需要执行具体任务时，优先查 role_skill_index，再查 evidence_index。
- 遇到真实安装、文件修改、网络访问、性能或兼容性问题时，转入 risk_card 和 boundaries.runtime_required。

### 任务路由

- **AI Skill / Agent 指令资产库**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`packages/claude/skills/brainstorming/SKILL.md`, `packages/claude/skills/debug-method/SKILL.md`, `packages/claude/skills/docs-builder/SKILL.md`, `packages/claude/skills/live-canvas/SKILL.md` 等 Claim：`clm_0001` supported 0.86
- **多宿主安装与分发**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`packages/claude/plugins/live-canvas-marketplace/.claude-plugin/marketplace.json`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/.claude-plugin/plugin.json` Claim：`clm_0002` supported 0.86
- **命令行启动或安装流程**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`README.md`, `packages/ampcode/commands/live-canvas/README.md`, `packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md` Claim：`clm_0003` supported 0.86

### 上下文规模

- 文件总数：218
- 重要文件覆盖：40/218
- 证据索引条目：78
- 角色 / Skill 条目：9

### 证据不足时的处理

- **missing_evidence**：说明证据不足，要求用户提供目标文件、README 段落或安装后验证记录；不要补全事实。
- **out_of_scope_request**：说明该任务超出当前 AI Context Pack 证据范围，并建议用户先查看 Human Manual 或真实安装后验证。
- **runtime_request**：给出安装前检查清单和命令来源，但不要替用户执行命令或声称已执行。
- **source_conflict**：同时展示冲突来源，标记为待核实，不要强行选择一个版本。

## Prompt Recipes

### 适配判断

- 目标：判断这个项目是否适合用户当前任务。
- 预期输出：适配结论、关键理由、证据引用、安装前可预览内容、必须安装后验证内容、下一步建议。

```text
请基于 liteagents-skill-installer 的 AI Context Pack，先问我 3 个必要问题，然后判断它是否适合我的任务。回答必须包含：适合谁、能做什么、不能做什么、是否值得安装、证据来自哪里。所有项目事实必须引用 evidence_refs、source_paths 或 claim_id。
```

### 安装前体验

- 目标：让用户在安装前感受核心工作流，同时避免把预览包装成真实能力或营销承诺。
- 预期输出：一段带边界标签的体验剧本、安装后验证清单和谨慎建议；不含真实运行承诺或强营销表述。

```text
请把 liteagents-skill-installer 当作安装前体验资产，而不是已安装工具或真实运行环境。

请严格输出四段：
1. 先问我 3 个必要问题。
2. 给出一段“体验剧本”：用 [安装前可预览]、[必须安装后验证]、[证据不足] 三种标签展示它可能如何引导工作流。
3. 给出安装后验证清单：列出哪些能力只有真实安装、真实宿主加载、真实项目运行后才能确认。
4. 给出谨慎建议：只能说“值得继续研究/试装”“先补充信息后再判断”或“不建议继续”，不得替项目背书。

硬性边界：
- 不要声称已经安装、运行、执行测试、修改文件或产生真实结果。
- 不要写“自动适配”“确保通过”“完美适配”“强烈建议安装”等承诺性表达。
- 如果描述安装后的工作方式，必须使用“如果安装成功且宿主正确加载 Skill，它可能会……”这种条件句。
- 体验剧本只能写成“示例台词/假设流程”：使用“可能会询问/可能会建议/可能会展示”，不要写“已写入、已生成、已通过、正在运行、正在生成”。
- Prompt Preview 不负责给安装命令；如用户准备试装，只能提示先阅读 Quick Start 和 Risk Card，并在隔离环境验证。
- 所有项目事实必须来自 supported claim、evidence_refs 或 source_paths；inferred/unverified 只能作风险或待确认项。

```

### 角色 / Skill 选择

- 目标：从项目里的角色或 Skill 中挑选最匹配的资产。
- 预期输出：候选角色或 Skill 列表，每项包含适用场景、证据路径、风险边界和是否需要安装后验证。

```text
请读取 role_skill_index，根据我的目标任务推荐 3-5 个最相关的角色或 Skill。每个推荐都要说明适用场景、可能输出、风险边界和 evidence_refs。
```

### 风险预检

- 目标：安装或引入前识别环境、权限、规则冲突和质量风险。
- 预期输出：环境、权限、依赖、许可、宿主冲突、质量风险和未知项的检查清单。

```text
请基于 risk_card、boundaries 和 quick_start_candidates，给我一份安装前风险预检清单。不要替我执行命令，只说明我应该检查什么、为什么检查、失败会有什么影响。
```

### 宿主 AI 开工指令

- 目标：把项目上下文转成一次对话开始前的宿主 AI 指令。
- 预期输出：一段边界明确、证据引用明确、适合复制给宿主 AI 的开工前指令。

```text
请基于 liteagents-skill-installer 的 AI Context Pack，生成一段我可以粘贴给宿主 AI 的开工前指令。这段指令必须遵守 not_runtime=true，不能声称项目已经安装、运行或产生真实结果。
```

## 角色 / Skill 索引

- 共索引 9 个角色 / Skill / 项目文档条目。

- **brainstorming**（skill）：Use when creating or developing, before writing code or implementation plans - refines rough ideas into fully-formed designs through collaborative questioning, alternative exploration, and incremental validation. Don't use during clear 'mechanical' processes 激活提示：当用户任务与“brainstorming”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`packages/claude/skills/brainstorming/SKILL.md`
- **debug-method**（skill）：Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes - four-phase framework root cause investigation, pattern analysis, hypothesis testing, implementation that ensures understanding before attempting solutions 激活提示：当用户任务与“debug-method”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`packages/claude/skills/debug-method/SKILL.md`
- **docs-builder**（skill）：Create or reorganize project documentation with structured /docs hierarchy 激活提示：当用户任务与“docs-builder”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`packages/claude/skills/docs-builder/SKILL.md`
- **live-canvas**（skill）：Conduct design interviews, generate UI variations, and collect live click-to-annotate feedback that streams into the session so edits land without leaving the browser. Use when the user wants rapid iterative UI refinement, not just batched feedback. 激活提示：当用户任务与“live-canvas”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`packages/claude/skills/live-canvas/SKILL.md`
- **skill-creator**（skill）：Guide for creating effective skills. This skill should be used when users want to create a new skill or update an existing skill that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations. 激活提示：当用户任务与“skill-creator”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`packages/claude/skills/skill-creator/SKILL.md`
- **tdd-flow**（skill）：Use when implementing any feature or bugfix, before writing implementation code - write the test first, watch it fail, write minimal code to pass; ensures tests actually verify behavior by requiring failure first 激活提示：当用户任务与“tdd-flow”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`packages/claude/skills/tdd-flow/SKILL.md`
- **test-traps**（skill）：Use when writing or changing tests, adding mocks, or tempted to add test-only methods to production code - prevents testing mock behavior, production pollution with test-only methods, and mocking without understanding dependencies 激活提示：当用户任务与“test-traps”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`packages/claude/skills/test-traps/SKILL.md`
- **trace-back**（skill）：Use when errors occur deep in execution and you need to trace back to find the original trigger - systematically traces bugs backward through call stack, adding instrumentation when needed, to identify source of invalid data or incorrect behavior 激活提示：当用户任务与“trace-back”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`packages/claude/skills/trace-back/SKILL.md`
- **verify-done**（skill）：Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always 激活提示：当用户任务与“verify-done”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`packages/claude/skills/verify-done/SKILL.md`

## 证据索引

- 共索引 78 条证据。

- **liteagents**（documentation）：AI development toolkit with 11 specialized agents and 20 commands per tool. Supports Claude, Opencode, Ampcode, and Droid. 证据：`CLAUDE.md`
- **Quick Start**（documentation）：AI development toolkit with 11 specialized agents and 18 commands per tool 证据：`README.md`
- **Live Canvas**（documentation）：Click-to-annotate UI design tool for Claude Code. Renders N variants of a component or page in your browser, you click an element and type what to change, and Claude edits the variant file while you watch in the browser. No window switching, no pasted JSON. 证据：`packages/ampcode/commands/live-canvas/README.md`
- **Global Claude Code CLI**（documentation）：Claude Code is a lightweight CLI tool that provides workflow automation commands. 证据：`packages/claude/CLAUDE.md`
- **live-canvas-channel**（documentation）：Claude Code channel plugin that bridges the Live Canvas browser overlay to a running Claude session. Each Save in the overlay becomes a notifications/claude/channel event — Claude reacts immediately without the user leaving the browser. 证据：`packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/README.md`
- **Live Canvas**（documentation）：Click-to-annotate UI design tool for Claude Code. Renders N variants of a component or page in your browser, you click an element and type what to change, and Claude edits the variant file while you watch in the browser. No window switching, no pasted JSON. 证据：`packages/claude/skills/live-canvas/README.md`
- **Live Canvas**（documentation）：Click-to-annotate UI design tool for Claude Code. Renders N variants of a component or page in your browser, you click an element and type what to change, and Claude edits the variant file while you watch in the browser. No window switching, no pasted JSON. 证据：`packages/droid/commands/live-canvas/README.md`
- **Live Canvas**（documentation）：Click-to-annotate UI design tool for Claude Code. Renders N variants of a component or page in your browser, you click an element and type what to change, and Claude edits the variant file while you watch in the browser. No window switching, no pasted JSON. 证据：`packages/opencode/command/live-canvas/README.md`
- **Package**（package_manifest）：{ "name": "liteagents", "version": "2.13.0", "description": "AI development toolkit with 11 specialized agents and 17 commands including live-canvas UI design with click-to-annotate feedback. Simple one-question installer for Claude, Opencode, Ampcode, and Droid.", "main": "index.js", "bin": { "liteagents": "./installer/cli.js", "liteag": "./cli.js" }, "scripts": { "test": "node tests/run-all-tests.js", "test-installer": "node tests/installer/integration.test.js", "install-interactive": "node installer/cli.js", "uninstall-tool": "node installer/cli.js --uninstall", "clean": "rm -rf tests/tmp/ && echo 'Cleaned test artifacts'", "clean:git": "git clean -fdx tests/tmp/", "validate": "node ./sc… 证据：`package.json`
- **Package**（package_manifest）：{ "name": "live-canvas-channel", "version": "0.5.0", "description": "MCP channel server: bridges Live Canvas HTTP overlay events into the Claude Code session.", "main": "server.js", "type": "commonjs", "engines": { "node": " =18" }, "scripts": { "start": "node server.js" }, "dependencies": { "@modelcontextprotocol/sdk": "^1.0.0" }, "license": "MIT" } 证据：`packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/package.json`
- **liteagents Installer Guide**（documentation）：Version : 2.8.1 Last Updated : 2026-05-24 证据：`docs/INSTALLER_GUIDE.md`
- **antigen gate "sleep" — PRD**（documentation）：Validation-gated hot memory for the liteagents /stash → /remember pipeline. No rule reaches always-loaded memory except through measured consequences — similarity/reflection proposes, consequence disposes. This is SkillOpt's validation-gate idea arXiv/microsoft SkillOpt, 2026 ported to the one domain where the reward cannot be replayed offline: the user's own future sessions. Governing rules: AGENT RULES.md POC-first, vanilla → stdlib → external, simple clever, surgical changes . friction's own doctrine applies throughout: precision over recall — a false hot rule loaded into every future session costs more than a missed faint one. Status legend: DECIDED settled · POC-GATED build only after… 证据：`docs/antigen-gate-prd.md`
- **Hot Memory — the stash → remember pipeline**（documentation）：Hot Memory — the stash → remember pipeline 证据：`docs/remember-README.md`
- **Global Droid CLI**（documentation）：Droid is a lightweight CLI tool that provides workflow automation commands. 证据：`packages/droid/AGENTS.md`
- **Global Opencode CLI**（documentation）：Opencode is a lightweight CLI tool that provides workflow automation commands. 证据：`packages/opencode/AGENTS.md`
- **Brainstorming Ideas Into Designs**（skill_instruction）：Help turn ideas into fully formed designs and specs through natural collaborative dialogue. 证据：`packages/claude/skills/brainstorming/SKILL.md`
- **Debug Method**（skill_instruction）：Random fixes waste time and create new bugs. Quick patches mask underlying issues. 证据：`packages/claude/skills/debug-method/SKILL.md`
- **Documentation Architecture Skill**（skill_instruction）：Create or reorganize /docs following a 5-tier hierarchy: 证据：`packages/claude/skills/docs-builder/SKILL.md`
- **Live Canvas Skill**（skill_instruction）：This skill implements a complete design exploration workflow: interview, generate variations, collect feedback, refine, preview, and finalize. 证据：`packages/claude/skills/live-canvas/SKILL.md`
- **Skill Creator**（skill_instruction）：This skill provides guidance for creating effective skills. 证据：`packages/claude/skills/skill-creator/SKILL.md`
- **TDD Flow**（skill_instruction）：APPLIES WHEN: - About to write new production code function, method, class - Implementing any feature or bug fix - Modifying behavior of existing code - Task metadata includes tdd: yes - User requests feature implementation 证据：`packages/claude/skills/tdd-flow/SKILL.md`
- **Trace Back**（skill_instruction）：Bugs often manifest deep in the call stack git init in wrong directory, file created in wrong location, database opened with wrong path . Your instinct is to fix where the error appears, but that's treating a symptom. 证据：`packages/claude/skills/trace-back/SKILL.md`
- **Verify Done**（skill_instruction）：APPLIES WHEN: - About to say: done, fixed, complete, passing, success - About to commit, push, create PR - About to mark task x 证据：`packages/claude/skills/verify-done/SKILL.md`
- **Marketplace**（structured_config）：{ "name": "live-canvas-marketplace", "description": "Live Canvas — push click-to-annotate feedback from the browser overlay into a Claude Code session.", "owner": { "name": "liteagents" }, "plugins": { "name": "live-canvas-channel", "source": "./plugins/live-canvas-channel", "description": "Channel plugin that bridges the Live Canvas overlay HTTP to the Claude session MCP notifications ." } } 证据：`packages/claude/plugins/live-canvas-marketplace/.claude-plugin/marketplace.json`
- **Plugin**（structured_config）：{ "name": "live-canvas-channel", "version": "0.5.0", "description": "Bridges the Live Canvas browser overlay to a Claude Code session. Each overlay Save becomes a claude/channel notification delivered over MCP.", "author": { "name": "liteagents" }, "channels": { "server": "live-canvas" } , "mcpServers": { "live-canvas": { "command": "node", "args": "${CLAUDE PLUGIN ROOT}/server.js" , "env": { "LIVE CANVAS PORT": "8788" } } } } 证据：`packages/claude/plugins/live-canvas-marketplace/plugins/live-canvas-channel/.claude-plugin/plugin.json`
- **License**（source_file）：Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ 证据：`LICENSE`
- **Test Traps**（skill_instruction）：Tests must verify real behavior, not mock behavior. Mocks are a means to isolate, not the thing being tested. 证据：`packages/claude/skills/test-traps/SKILL.md`
- **Changelog**（documentation）：All notable changes to liteagents will be documented in this file. 证据：`CHANGELOG.md`
- **Workflow**（documentation）：You are an expert Product Manager creating clear, actionable PRDs for junior developers. 证据：`packages/ampcode/agents/1-create-prd.md`
- **CRITICAL BEHAVIOR**（documentation）：You are an expert Technical Program Manager translating PRDs into precise, actionable task lists for non-technical users, accounting for existing codebase patterns. 证据：`packages/ampcode/agents/2-generate-tasks.md`
- **RULES**（documentation）：You are an implementation agent executing tasks from a provided task list. 证据：`packages/ampcode/agents/3-process-task-list.md`
- **Session Start**（documentation）：You are a router. You don't do work—you match intent to agents/workflows, spawn with minimal context, and track state. 证据：`packages/ampcode/agents/orchestrator.md`
- **── Add your own global sessions root at the TOP so it is checked first ──**（documentation）：Run friction analysis, then consolidate session stashes + friction antigens into a single project-local MEMORY.md, and inject into AGENT.md. Friction runs automatically best-effort — there is no separate /friction command. 证据：`packages/ampcode/commands/remember.md`
- **Stash with auto-generated name**（documentation）：Save session context for compaction recovery or handoffs. 证据：`packages/ampcode/commands/stash.md`
- **Workflow**（documentation）：You are an expert Product Manager creating clear, actionable PRDs for junior developers. 证据：`packages/claude/agents/1-create-prd.md`
- **CRITICAL BEHAVIOR**（documentation）：You are an expert Technical Program Manager translating PRDs into precise, actionable task lists for non-technical users, accounting for existing codebase patterns. 证据：`packages/claude/agents/2-generate-tasks.md`
- **RULES**（documentation）：You are an implementation agent executing tasks from a provided task list. 证据：`packages/claude/agents/3-process-task-list.md`
- **Session Start**（documentation）：You are a router. You don't do work—you match intent to agents/workflows, spawn with minimal context, and track state. 证据：`packages/claude/agents/orchestrator.md`
- **── Add your own global sessions root at the TOP so it is checked first ──**（documentation）：Run friction analysis, then consolidate session stashes + friction antigens into a single project-local MEMORY.md, and inject into CLAUDE.md. Friction runs automatically best-effort — there is no separate /friction command. 证据：`packages/claude/commands/remember.md`
- **Stash with auto-generated name**（documentation）：Save session context for compaction recovery or handoffs. 证据：`packages/claude/commands/stash.md`
- **── Add your own global sessions root at the TOP so it is checked first ──**（documentation）：Run friction analysis, then consolidate session stashes + friction antigens into a single project-local MEMORY.md, and inject into AGENTS.md. Friction runs automatically best-effort — there is no separate /friction command. 证据：`packages/droid/commands/remember.md`
- **Stash with auto-generated name**（documentation）：Save session context for compaction recovery or handoffs. 证据：`packages/droid/commands/stash.md`
- **Workflow**（documentation）：You are an expert Product Manager creating clear, actionable PRDs for junior developers. 证据：`packages/droid/droids/1-create-prd.md`
- **CRITICAL BEHAVIOR**（documentation）：You are an expert Technical Program Manager translating PRDs into precise, actionable task lists for non-technical users, accounting for existing codebase patterns. 证据：`packages/droid/droids/2-generate-tasks.md`
- **RULES**（documentation）：You are an implementation agent executing tasks from a provided task list. 证据：`packages/droid/droids/3-process-task-list.md`
- **Session Start**（documentation）：You are a router. You don't do work—you match intent to agents/workflows, spawn with minimal context, and track state. 证据：`packages/droid/droids/orchestrator.md`
- **Workflow**（documentation）：You are an expert Product Manager creating clear, actionable PRDs for junior developers. 证据：`packages/opencode/agent/1-create-prd.md`
- **CRITICAL BEHAVIOR**（documentation）：You are an expert Technical Program Manager translating PRDs into precise, actionable task lists for non-technical users, accounting for existing codebase patterns. 证据：`packages/opencode/agent/2-generate-tasks.md`
- **RULES**（documentation）：You are an implementation agent executing tasks from a provided task list. 证据：`packages/opencode/agent/3-process-task-list.md`
- **Session Start**（documentation）：You are a router. You don't do work—you match intent to agents/workflows, spawn with minimal context, and track state. 证据：`packages/opencode/agent/orchestrator.md`
- **── Add your own global sessions root at the TOP so it is checked first ──**（documentation）：Run friction analysis, then consolidate session stashes + friction antigens into a single project-local MEMORY.md, and inject into AGENTS.md. Friction runs automatically best-effort — there is no separate /friction command. 证据：`packages/opencode/command/remember.md`
- **Stash with auto-generated name**（documentation）：Save session context for compaction recovery or handoffs. 证据：`packages/opencode/command/stash.md`
- **Cli**（source_file）：class InteractiveInstaller ⋮---- async askInstallOrUninstall ⋮---- async run ⋮---- async runUninstall ⋮---- const rimraf = dir = ⋮---- findBackups targetPath ⋮---- async selectToolsCheckbox tools ⋮---- const renderList = = ⋮---- console.log '' ; // Empty line at bottom ⋮---- // Initial render ⋮---- const onKeypress = key = ⋮---- async handleFatalError error ⋮---- console.log '' ; // Add spacing ⋮---- // Categorize the error and provide appropriate guidance ⋮---- // Exit with error code ⋮---- / Categorize errors and provide actionable advice @param {Error} error - The error to categorize @returns {object} Object with type, advice array, and technical details / categorizeError error ⋮---- //… 证据：`installer/cli.js`
- **Installation Engine**（source_file）：class InstallationEngine ⋮---- async installMultipleTools variant, tools, paths, progressCallback = null, resume = false ⋮---- async installPackage toolId, variant, targetPath, progressCallback = null ⋮---- async copySelectedFiles sourceBase, targetPath, packageContents, progressCallback = null ⋮---- const getFileSize = async filePath = ⋮---- const collectDirectoryFiles = async dirPath, relativePath = '', skipNodeModules = false = ⋮---- async copyDirectory source, target ⋮---- async generateManifest toolId, variant, targetPath ⋮---- const extractAgentName = fullPath = ⋮---- const extractSkillName = fullPath = ⋮---- async createBackup targetPath ⋮---- async rollbackInstallation toolId, targe… 证据：`installer/installation-engine.js`
- **Package Manager**（source_file）：class PackageManager ⋮---- async loadVariantConfig toolId ⋮---- async getVariantMetadata toolId, variant ⋮---- async selectVariantContent toolId, variant, availableContent ⋮---- const selectItems = category, variantSelection, availableItems, skipMissing = false = ⋮---- getAvailableVariants toolId ⋮---- async getPackageContents toolId, variant ⋮---- async getAvailableContent packageDir ⋮---- const getItemsInDir = async dir, isAgentDir = false = ⋮---- async countFiles dir ⋮---- async function traverse currentDir ⋮---- async getPackageSize toolId, variant ⋮---- const calculatePathSize = async itemPath = ⋮---- formatBytes bytes ⋮---- async validatePackage toolId, variant ⋮---- const checkConten… 证据：`installer/package-manager.js`
- **Path Manager**（source_file）：class PathManager ⋮---- expandPath pathStr ⋮---- sanitizePath pathStr ⋮---- async validatePath pathStr ⋮---- async getDiskSpace pathStr ⋮---- async checkExistingInstallation pathStr ⋮---- getDefaultPath toolId ⋮---- normalizePathForDisplay pathStr 证据：`installer/path-manager.js`
- **Report Template**（source_file）：class ReportTemplate ⋮---- generateReport installationData ⋮---- // Header ⋮---- // Summary Section ⋮---- // Tool Details Section ⋮---- // System Information Section ⋮---- // Errors Section if any ⋮---- // Warnings Section if any ⋮---- // Footer ⋮---- / Save report to file @param {string} reportContent - Generated report content @returns {Promise } Path where report was saved / async saveReport reportContent ⋮---- async createAndSaveReport installationData ⋮---- getReportPath ⋮---- generateHeader ⋮---- generateSummary summary ⋮---- generateToolDetails tools ⋮---- generateSystemInfo ⋮---- generateErrorsSection errors ⋮---- generateWarningsSection warnings ⋮---- generateFooter ⋮---- formatDur… 证据：`installer/report-template.js`
- **Friction**（source_file）：function loadConfig ⋮---- function parseJsonl raw, source ⋮---- function parseISODate s ⋮---- function formatDuration minutes ⋮---- function globFiles dir, pattern, recursive ⋮---- function listDirs dir ⋮---- // ============================================================================= // FRICTION ANALYZE - derive session name // ============================================================================= ⋮---- function deriveSessionName sessionFile, metadata ⋮---- function extractToolNameFromResult result ⋮---- function looksLikeTerminalPaste text ⋮---- function extractSignals sessionFile ⋮---- content = text; // Fall through to user message handling below ⋮---- // Handle single dict t… 证据：`packages/ampcode/commands/remember/friction.js`
- **Friction**（source_file）：function loadConfig ⋮---- function parseJsonl raw, source ⋮---- function parseISODate s ⋮---- function formatDuration minutes ⋮---- function globFiles dir, pattern, recursive ⋮---- function listDirs dir ⋮---- // ============================================================================= // FRICTION ANALYZE - derive session name // ============================================================================= ⋮---- function deriveSessionName sessionFile, metadata ⋮---- function extractToolNameFromResult result ⋮---- function looksLikeTerminalPaste text ⋮---- function extractSignals sessionFile ⋮---- content = text; // Fall through to user message handling below ⋮---- // Handle single dict t… 证据：`packages/claude/commands/remember/friction.js`
- **Friction**（source_file）：function loadConfig ⋮---- function parseJsonl raw, source ⋮---- function parseISODate s ⋮---- function formatDuration minutes ⋮---- function globFiles dir, pattern, recursive ⋮---- function listDirs dir ⋮---- // ============================================================================= // FRICTION ANALYZE - derive session name // ============================================================================= ⋮---- function deriveSessionName sessionFile, metadata ⋮---- function extractToolNameFromResult result ⋮---- function looksLikeTerminalPaste text ⋮---- function extractSignals sessionFile ⋮---- content = text; // Fall through to user message handling below ⋮---- // Handle single dict t… 证据：`packages/droid/commands/remember/friction.js`
- 其余 18 条证据见 `AI_CONTEXT_PACK.json` 或 `EVIDENCE_INDEX.json`。

## 宿主 AI 必须遵守的规则

- **把本资产当作开工前上下文，而不是运行环境。**：AI Context Pack 只包含证据化项目理解，不包含目标项目的可执行状态。 证据：`CLAUDE.md`, `README.md`, `packages/ampcode/commands/live-canvas/README.md`
- **回答用户时区分可预览内容与必须安装后才能验证的内容。**：安装前体验的消费者价值来自降低误装和误判，而不是伪装成真实运行。 证据：`CLAUDE.md`, `README.md`, `packages/ampcode/commands/live-canvas/README.md`

## 用户开工前应该回答的问题

- 你准备在哪个宿主 AI 或本地环境中使用它？
- 你只是想先体验工作流，还是准备真实安装？
- 你最在意的是安装成本、输出质量、还是和现有规则的冲突？

## 验收标准

- 所有能力声明都能回指到 evidence_refs 中的文件路径。
- AI_CONTEXT_PACK.md 没有把预览包装成真实运行。
- 用户能在 3 分钟内看懂适合谁、能做什么、如何开始和风险边界。

---

## Doramagic Context Augmentation

下面内容用于强化 Repomix/AI Context Pack 主体。Human Manual 只提供阅读骨架；踩坑日志会被转成宿主 AI 必须遵守的工作约束。

## Human Manual 骨架

使用规则：这里只是项目阅读路线和显著性信号，不是事实权威。具体事实仍必须回到 repo evidence / Claim Graph。

宿主 AI 硬性规则：
- 不得把页标题、章节顺序、摘要或 importance 当作项目事实证据。
- 解释 Human Manual 骨架时，必须明确说它只是阅读路线/显著性信号。
- 能力、安装、兼容性、运行状态和风险判断必须引用 repo evidence、source path 或 Claim Graph。

- **项目总览与架构（Project Overview & Architecture）**：importance `high`
  - source_paths: README.md, package.json, cli.js, postinstall.js, CHANGELOG.md
- **11 个智能体与 22 个命令/技能目录（Agents & Commands Catalog）**：importance `high`
  - source_paths: packages/claude/agents/orchestrator.md, packages/claude/agents/1-create-prd.md, packages/claude/agents/2-generate-tasks.md, packages/claude/agents/3-process-task-list.md, packages/claude/commands/remember.md
- **热记忆管线与摩擦抗原系统（Hot Memory & Friction Antigen Pipeline，v2.9.0 重点）**：importance `high`
  - source_paths: docs/remember-README.md, docs/antigen-gate-prd.md, packages/claude/commands/remember.md, packages/claude/commands/remember/friction.js, packages/claude/commands/stash.md
- **多工具分发、安装器与扩展机制（Multi-Tool Distribution, Installer & Extensibility）**：importance `high`
  - source_paths: docs/INSTALLER_GUIDE.md, installer/cli.js, installer/installation-engine.js, installer/package-manager.js, installer/path-manager.js

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `3ea5142785d077b71ea2eadb44057557c969b886`
- inspected_files: `README.md`, `package.json`, `docs/INSTALLER_GUIDE.md`, `docs/antigen-gate-prd.md`, `docs/remember-README.md`, `packages/ampcode/AGENT.md`, `packages/ampcode/agents/1-create-prd.md`, `packages/ampcode/agents/2-generate-tasks.md`, `packages/ampcode/agents/3-process-task-list.md`, `packages/ampcode/agents/code-developer.md`, `packages/ampcode/agents/context-builder.md`, `packages/ampcode/agents/feature-planner.md`, `packages/ampcode/agents/market-researcher.md`, `packages/ampcode/agents/orchestrator.md`, `packages/ampcode/agents/quality-assurance.md`, `packages/ampcode/agents/system-architect.md`, `packages/ampcode/agents/ui-designer.md`, `packages/ampcode/commands/brainstorming.md`, `packages/ampcode/commands/debug-method.md`, `packages/ampcode/commands/diff-review.md`

宿主 AI 硬性规则：
- 没有 repo_clone_verified=true 时，不得声称已经读过源码。
- 没有 repo_inspection_verified=true 时，不得把 README/docs/package 文件判断写成事实。
- 没有 quick_start_verified=true 时，不得声称 Quick Start 已跑通。

## Doramagic Pitfall Constraints / 踩坑约束

这些规则来自 Doramagic 发现、验证或编译过程中的项目专属坑点。宿主 AI 必须把它们当作工作约束，而不是普通说明文字。

### Constraint 1: 仓库名和安装名不一致

- Trigger: 仓库名 `liteagents-skill-installer` 与安装入口 `liteagents` 不完全一致。
- Host AI rule: 在 npm/PyPI/GitHub 上确认包名映射和官方 README 说明。
- Why it matters: 用户照着仓库名搜索包或照着包名找仓库时容易走错入口。
- Evidence: identity.distribution | https://github.com/hamr0/liteagents | repo=liteagents-skill-installer; install=liteagents
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 2: 可能修改宿主 AI 配置

- Trigger: 项目面向 Claude/Cursor/Codex/Gemini/OpenCode 等宿主，或安装命令涉及用户配置目录。
- Host AI rule: 列出会写入的配置文件、目录和卸载/回滚步骤。
- Why it matters: 安装可能改变本机 AI 工具行为，用户需要知道写入位置和回滚方法。
- Evidence: capability.host_targets | https://github.com/hamr0/liteagents | host_targets=mcp_host, claude_code, claude
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 3: 能力判断依赖假设

- Trigger: README/documentation is current enough for a first validation pass.
- Host AI rule: 将假设转成下游验证清单。
- Why it matters: 假设不成立时，用户拿不到承诺的能力。
- Evidence: capability.assumptions | https://github.com/hamr0/liteagents | README/documentation is current enough for a first validation pass.
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 4: 维护活跃度未知

- Trigger: 未记录 last_activity_observed。
- Host AI rule: 补 GitHub 最近 commit、release、issue/PR 响应信号。
- Why it matters: 新项目、停更项目和活跃项目会被混在一起，推荐信任度下降。
- Evidence: evidence.maintainer_signals | https://github.com/hamr0/liteagents | last_activity_observed missing
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

- Trigger: no_demo
- Evidence: downstream_validation.risk_items | https://github.com/hamr0/liteagents | no_demo; severity=medium
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 6: 存在评分风险

- Trigger: no_demo
- Why it matters: 风险会影响是否适合普通用户安装。
- Evidence: risks.scoring_risks | https://github.com/hamr0/liteagents | no_demo; severity=medium
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 7: issue/PR 响应质量未知

- Trigger: issue_or_pr_quality=unknown。
- Host AI rule: 抽样最近 issue/PR，判断是否长期无人处理。
- Why it matters: 用户无法判断遇到问题后是否有人维护。
- Evidence: evidence.maintainer_signals | https://github.com/hamr0/liteagents | issue_or_pr_quality=unknown
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 8: 发布节奏不明确

- Trigger: release_recency=unknown。
- Host AI rule: 确认最近 release/tag 和 README 安装命令是否一致。
- Why it matters: 安装命令和文档可能落后于代码，用户踩坑概率升高。
- Evidence: evidence.maintainer_signals | https://github.com/hamr0/liteagents | release_recency=unknown
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。