lifeos 1.0.0

package/assets/skills/research/references/execution-agent-prompt.en.md

@@ -0,0 +1,166 @@
+---
+name: execution-agent-prompt
+description: Execution Agent prompt for the Research skill
+role: execution
+parent_skill: research
+---
+
+# Research Execution Agent Instructions
+
+> Path logical names (e.g., `{research directory}`, `{drafts directory}`) are resolved by the Orchestrator from `lifeos.yaml` and injected into context. See the main skill file `research/SKILL.md` for the mapping.
+
+> This file is read by the `research/SKILL.md` Orchestrator after the user confirms the plan, and used as the complete prompt for the Task tool.
+> Replace `[plan file path]` with the actual plan file path when using.
+
+---
+
+Execute the research plan at the following path: [plan file path]
+
+## Step 1: Read the Plan File in Full
+
+Pay attention to the following key fields:
+
+- Trigger mode (FILE MODE or TOPIC MODE)
+- Local draft materials (listed file paths)
+- Clarification question answers (user's knowledge level and method preference)
+- Expert persona (file path, applicability mode, special format requirements)
+
+**Persona application rules (based on applicability mode):**
+
+| Applicability mode        | Execution rules                                                                                                                                                                                                                                                                  |
+| ------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `Full apply`              | Read the complete persona file; use the persona's Analytical Framework as the structural skeleton; replace default chapters with the persona's Output Format (always retain: frontmatter, core insights from drafts, related reading, references); enforce the persona's Constraints and Guardrails; match the persona's Interaction Style tone |
+| `Reference apply`         | Read the persona file, borrow its analysis framework to enrich the analysis, but retain the default chapter structure below                                                                                                                                                       |
+| `Not applicable` or `Not found` | Use the default chapter structure below                                                                                                                                                                                                                                    |
+
+## Step 2: Read Local Draft Materials
+
+- Read the full content of each file listed in "Local Draft Materials" one by one
+- Treat these files as **first-hand sources** (representing the user's original thinking)
+- **FILE MODE**: the specified draft file is the core anchor; all research expands from it
+- **TOPIC MODE**: draft files serve as supplementary background
+
+## Step 3: External Research
+
+- Use WebSearch to retrieve current information, official documentation, authoritative sources
+- Use WebFetch to read documentation pages
+- Cross-validate local draft insights with external sources
+- **When WebSearch returns nothing**: rely on local drafts, note limitations in the report
+- **When WebFetch fails**: mark in "References" as "(link inaccessible, for reference only)"
+
+## Step 4: Write the Research Report
+
+Path: `{research directory}/Domain/Topic/Topic.md`
+
+> ⚠️ `/research` must never create files under `{knowledge directory}/` — that is the responsibility of `/knowledge`.
+
+Adjust depth and style based on user's knowledge level and method preference.
+
+**Default chapter structure** (used when persona applicability mode is not `Full apply`):
+
+```markdown
+---
+title: "Topic"
+type: research
+created: "YYYY-MM-DD"
+domain: "[[DomainName]]"
+status: complete
+tags: [research]
+aliases: []
+---
+
+## Overview
+
+(Conclusion first: what it is, what problem it solves, scope of applicability)
+
+## Core Insights from Drafts
+
+(Only when local drafts exist: integrate the user's original ideas, cite source files. Omit this section if no drafts)
+
+## Core Concept Framework
+
+(Glossary/modules/components/key mechanisms; may use [[wikilink]] to point to existing knowledge/concepts, but do not create new {knowledge directory} files)
+
+## How It Works
+
+(Necessary technical details)
+
+## Examples
+
+(Code/scenarios/step-by-step procedures, if applicable)
+
+## Best Practices
+
+## Common Pitfalls
+
+## Related Reading
+
+(Wikilinks only: related research/projects/knowledge notes within the Vault)
+
+## References
+
+(External links: docs/articles/videos)
+```
+
+**Math formula format specification** (when Math persona is active, or when the report contains formulas):
+
+- Inline formulas: `$formula$` (single dollar signs)
+- Display formulas: `$$formula$$` (double dollar signs, each on its own line)
+- Multi-line formulas: `$$\begin{aligned}...\end{aligned}$$`
+- Do not use `\(...\)` or `\[...\]` (Obsidian does not render these by default)
+- Do not skip key derivation steps with "obviously" or "easily verified"
+
+## Step 5: Create Visualization Map (For Complex Topics)
+
+Path: `{research directory}/Domain/Topic/Topic_Map.canvas`
+
+## Step 6: Create Examples (If Applicable)
+
+Path: `{research directory}/Domain/Topic/examples/`
+
+## Step 7: Update Draft Status (Critical)
+
+For each draft file used from "Local Draft Materials":
+
+- Update the `status` in its frontmatter to `researched`
+- This marks the draft as processed, allowing `/archive` to identify and archive it
+
+## Step 8: Update Today's Diary
+
+If `{diary directory}/YYYY-MM-DD.md` exists, append a brief research summary. Skip this step if the diary file does not exist.
+
+## Step 9: Update Plan Status (Critical)
+
+- After the research report is complete, update the plan file frontmatter `status` to `done`
+- Keep the plan file in `{plans directory}/`
+- `/archive` later moves plans with `status: done` into `{system directory}/{archived plans subdirectory}/`
+
+---
+
+## Completion Report
+
+After completion, report in English:
+
+```
+## Research Complete: [Topic]
+
+**Created:**
+- Research report: [[Topic]] ({research directory}/Domain/Topic/)
+- Examples: [N] files (if any)
+- Visualization: [Yes/No] (if any)
+
+**Integrated draft sources:**
+- [List draft files used, or "None"] → status updated to researched
+
+**Plan status:** {plans directory}/Plan_YYYY-MM-DD_Research_Topic.md → `status: done` (waiting for `/archive` to move it into `{system directory}/{archived plans subdirectory}/`)
+
+**Key takeaways:**
+1. [Takeaway 1]
+2. [Takeaway 2]
+3. [Takeaway 3]
+
+**Next steps:**
+- [ ] To distill into reusable knowledge points: use /knowledge to extract from authoritative sources into {knowledge directory}/{notes subdirectory} or {wiki subdirectory}
+- [ ] Consolidate through practice/projects (if applicable)
+- [ ] Run /archive to clean up processed drafts
+```

package/assets/skills/research/references/execution-agent-prompt.zh.md

@@ -0,0 +1,166 @@
+---
+name: execution-agent-prompt
+description: Research 技能的执行 Agent 提示词
+role: execution
+parent_skill: research
+---
+
+# 研究执行 Agent 指令
+
+> 路径逻辑名（如 `{研究目录}`、`{草稿目录}`）由 Orchestrator 从 `lifeos.yaml` 解析后注入上下文。映射关系见主技能文件 `research/SKILL.md` 的配置块。
+
+> 此文件由 `research/SKILL.md` 的 Orchestrator 在用户确认计划后读取，作为 Task 工具的完整 prompt 使用。
+> 使用时将 `[plan file path]` 替换为实际计划文件路径。
+
+---
+
+执行以下路径的研究计划：[计划文件路径]
+
+## 步骤一：完整读取计划文件
+
+注意以下关键字段：
+
+- 触发方式（FILE MODE or TOPIC MODE）
+- 本地草稿资料（列出的文件路径）
+- 澄清问题回答（用户的知识水平和方法偏好）
+- 专家人格（文件路径、适用模式、特殊格式要求）
+
+**人格应用规则（基于适用模式）：**
+
+| 适用模式             | 执行规则                                                                                                                                                                                                                                                                  |
+| -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `全量应用`           | 读取完整 persona 文件；用 persona 的 Analytical Framework 作为结构骨架；用 persona 的 Output Format 替换默认章节（始终保留：frontmatter、来自草稿的核心洞察、相关阅读、参考资源）；强制执行 persona 的 Constraints and Guardrails；匹配 persona 的 Interaction Style 语气 |
+| `参考应用`           | 读取 persona 文件，借用其分析框架丰富分析，但保留下方默认章节结构                                                                                                                                                                                                         |
+| `不适用` 或 `未找到` | 使用下方默认章节结构                                                                                                                                                                                                                                                      |
+
+## 步骤二：读取本地草稿素材
+
+- 逐一读取「本地草稿资料」中列出的每个文件的完整内容
+- 将这些文件作为**第一手来源**（代表用户的原始思考）
+- **FILE MODE**：指定的草稿文件是核心锚点，所有研究从它展开
+- **TOPIC MODE**：草稿文件作为补充背景
+
+## 步骤三：外部研究
+
+- 使用 WebSearch 检索当前信息、官方文档、权威来源
+- 使用 WebFetch 读取文档页面
+- 将本地草稿洞察与外部来源交叉验证
+- **WebSearch 无结果时**：依赖本地草稿，在报告中注明局限性
+- **WebFetch 失败时**：在「参考资源」标注"(链接无法访问，仅供参考)"
+
+## 步骤四：撰写研究报告
+
+路径：`{研究目录}/Domain/Topic/Topic.md`
+
+> ⚠️ `/research` 绝不在 `{知识目录}/` 下创建文件，那是 `/knowledge` 的职责。
+
+根据用户知识水平和方法偏好调整深度与风格。
+
+**默认章节结构**（当 persona 适用模式不是`全量应用`时使用）：
+
+```markdown
+---
+title: "Topic"
+type: research
+created: "YYYY-MM-DD"
+domain: "[[DomainName]]"
+status: complete
+tags: [research]
+aliases: []
+---
+
+## 概述
+
+（结论先行：这是什么、解决什么问题、适用边界）
+
+## 来自草稿的核心洞察
+
+（仅当有本地草稿时：整合用户原始想法，标注来源文件。无草稿时省略此节）
+
+## 核心概念框架
+
+（术语表/模块/组件/关键机制；可用 [[wikilink]] 指向已有知识/概念，但不新建 {知识目录} 文件）
+
+## 工作原理
+
+（必要的技术细节）
+
+## 示例
+
+（代码/场景/操作步骤，如适用）
+
+## 最佳实践
+
+## 常见陷阱
+
+## 相关阅读
+
+（仅 wikilinks：Vault 内相关研究/项目/知识笔记）
+
+## 参考资源
+
+（外部链接：docs/articles/videos）
+```
+
+**数学公式格式规范**（当 Math persona 激活，或报告含公式时）：
+
+- 行内公式：`$formula$`（单美元符号）
+- 独立公式：`$$formula$$`（双美元符号，各占一行）
+- 多行公式：`$$\begin{aligned}...\end{aligned}$$`
+- 禁止使用 `\(...\)` 或 `\[...\]`（Obsidian 默认不渲染）
+- 禁止用"显然"或"容易验证"跳过关键推导步骤
+
+## 步骤五：创建可视化地图（复杂主题时）
+
+路径：`{研究目录}/Domain/Topic/Topic_Map.canvas`
+
+## 步骤六：创建示例（如适用）
+
+路径：`{研究目录}/Domain/Topic/examples/`
+
+## 步骤七：更新草稿状态（关键）
+
+对「本地草稿资料」中每个已使用的草稿文件：
+
+- 将 frontmatter 中的 `status` 更新为 `researched`
+- 这标记该草稿已被处理，使 `/archive` 可识别并归档
+
+## 步骤八：更新今日日记
+
+若 `{日记目录}/YYYY-MM-DD.md` 存在，追加简短研究摘要。若日记文件不存在则跳过此步骤。
+
+## 步骤九：更新计划状态（关键）
+
+- 研究报告完成后，将计划文件的 frontmatter 中 `status` 更新为 `done`
+- 保持计划文件仍位于 `{计划目录}/`
+- 后续由 `/archive` 统一将 `status: done` 的计划移动到 `{系统目录}/{归档计划子目录}/`
+
+---
+
+## 完成报告
+
+完成后用中文汇报：
+
+```
+## 研究完成: [Topic]
+
+**已创建:**
+- 研究报告: [[Topic]]（{研究目录}/Domain/Topic/）
+- 示例: [N] 个文件（如有）
+- 可视化: [是/否]（如有）
+
+**整合的草稿来源:**
+- [列出使用的草稿文件，或"无"] → 状态已更新为 researched
+
+**计划状态:** {计划目录}/Plan_YYYY-MM-DD_Research_Topic.md → `status: done`（待 `/archive` 归档到 `{系统目录}/{归档计划子目录}/`）
+
+**核心要点:**
+1. [要点1]
+2. [要点2]
+3. [要点3]
+
+**下一步:**
+- [ ] 如需沉淀为可复用知识点：用 /knowledge 将权威资料提取为 {知识目录}/{笔记子目录} 或 {百科子目录}
+- [ ] 通过实践/项目巩固（如适用）
+- [ ] 运行 /archive 清理已处理的草稿
+```

package/assets/skills/research/references/planning-agent-prompt.en.md

@@ -0,0 +1,129 @@
+---
+name: planning-agent-prompt
+description: Planning Agent prompt for the Research skill
+role: planning
+parent_skill: research
+---
+
+# Research Planning Agent Instructions
+
+> Path logical names (e.g., `{research directory}`, `{drafts directory}`) are resolved by the Orchestrator from `lifeos.yaml` and injected into context. See the main skill file `research/SKILL.md` for the mapping.
+
+> This file is read by the `research/SKILL.md` Orchestrator and used as the complete prompt for the Task tool.
+> Replace `[user's input]` with the user's actual input when using.
+
+---
+
+Create a research plan for the following: [user's input]
+
+Execute the following steps:
+
+## Step 1: Identify Trigger Mode
+
+- If the input is a file path or filename → **FILE MODE**
+  - Read the specified file from `{drafts directory}/`
+  - Extract core topics, questions, and key insights from the file
+  - This file is the primary source, treated as first-hand material
+- If the input is a topic/keyword → **TOPIC MODE**
+  - Use the input directly as the research topic
+
+## Step 2: Scan Local Drafts (`{drafts directory}/`)
+
+- List all `status: pending` files in `{drafts directory}/`
+- **TOPIC MODE**: identify files related to the topic via filenames and content keywords
+- **FILE MODE**: identify other files related to the specified file's topic
+- List the related draft files found (for the Execution Agent to use as local first-hand material)
+
+## Step 3: Check Existing Research
+
+- Search `{research directory}/` to avoid duplication
+- If related research exists: flag for update rather than new creation
+
+## Step 4: Determine Knowledge Domain
+
+- Infer the relevant domain (e.g., SoftwareEngineering, Finance, Health, ProductDesign, AI, Math, Art, History)
+- If undetermined: mark as `TBD`, the Orchestrator will ask the user
+
+## Step 5: Match Expert Persona
+
+- List all `.md` files in `{system directory}/提示词/`
+- For each file, read its **frontmatter** (`id`, `aliases`, `tags`) and the **Domain Coverage** / **Core Responsibilities** sections
+- Compare the research topic (from Step 4) against each persona's domain coverage to find the best match
+- If multiple personas could apply, pick the one whose domain is most specific to the topic
+- If no persona's domain overlaps with the research topic, note "No match found, using general research mode"
+
+**Applicability mode determination:**
+
+- `Full apply`: The research topic falls squarely within the persona's core domain — replace default chapter structure entirely with the persona's Output Format
+- `Reference apply`: The persona's analytical framework is relevant but the topic is not its core focus — borrow the analysis framework only, retain default chapters
+- `Not applicable`: No persona matches — use default report structure
+
+## Step 6: Create Plan File
+
+Path: `{plans directory}/Plan_YYYY-MM-DD_Research_Topic.md`
+
+```markdown
+---
+title: "Plan: [Topic]"
+type: plan
+status: active
+created: "YYYY-MM-DD"
+source: research
+topic: "[Topic]"
+tags: [plan, research]
+aliases: []
+---
+
+# Research Plan: [Topic]
+
+## Trigger Mode
+
+[FILE MODE: {drafts directory}/filename.md | TOPIC MODE: topic keywords]
+
+## Research Objective
+
+[What the user will understand after this research is complete]
+
+## Local Draft Materials
+
+[List related file paths in {drafts directory}/, or "No related drafts found"]
+
+## Existing Research
+
+[List related existing research reports in {research directory}/, or "Not found"]
+
+## Related Domain
+
+[Domain name, or TBD]
+
+## Expert Persona
+
+- File path: [{system directory}/提示词/XXX.md, or "Not found"]
+- Applicability reason: [Why this persona fits the research topic, one sentence]
+- Applicability mode: [Full apply | Reference apply | Not applicable]
+- Special format requirements: [e.g., LaTeX formulas, historical source citation format, or "None"]
+
+## Research Strategy
+
+[ ] Read local drafts as first-hand material (if available)
+[ ] Search official documentation/authoritative sources
+[ ] Form conceptual framework (glossary, module relationships, key mechanisms)
+[ ] Find practical examples and use cases (if applicable)
+[ ] Summarize best practices and common pitfalls
+[ ] Output actionable next-step recommendations
+
+## Output Structure (output only to {research directory}/)
+
+- Main note (research report): {research directory}/Domain/Topic/Topic.md
+- Examples/resources (optional): {research directory}/Domain/Topic/examples/
+- Visualization (optional): {research directory}/Domain/Topic/Topic_Map.canvas
+
+## Clarification Question Answers
+
+Knowledge level: [To be filled by user]
+Method preference: [To be filled by user]
+```
+
+## Step 7: Return Result
+
+Return the path to the plan file.

package/assets/skills/research/references/planning-agent-prompt.zh.md

@@ -0,0 +1,129 @@
+---
+name: planning-agent-prompt
+description: Research 技能的规划 Agent 提示词
+role: planning
+parent_skill: research
+---
+
+# 研究规划 Agent 指令
+
+> 路径逻辑名（如 `{研究目录}`、`{草稿目录}`）由 Orchestrator 从 `lifeos.yaml` 解析后注入上下文。映射关系见主技能文件 `research/SKILL.md` 的配置块。
+
+> 此文件由 `research/SKILL.md` 的 Orchestrator 读取，作为 Task 工具的完整 prompt 使用。
+> 使用时将 `[user's input]` 替换为用户实际输入内容。
+
+---
+
+为以下内容制定研究计划：[用户输入]
+
+按以下步骤执行：
+
+## 步骤一：识别触发模式
+
+- 如果输入是文件路径或文件名 → **FILE MODE**
+  - 从 `{草稿目录}/` 读取指定文件
+  - 提取文件中的核心主题、问题和关键观点
+  - 该文件为主要来源，视为第一手资料
+- 如果输入是主题/关键词 → **TOPIC MODE**
+  - 直接将输入作为研究主题
+
+## 步骤二：扫描本地草稿（`{草稿目录}/`）
+
+- 列出 `{草稿目录}/` 中所有 `status: pending` 的文件
+- **TOPIC MODE**：通过文件名和内容关键词识别与主题相关的文件
+- **FILE MODE**：识别与指定文件主题相关的其他文件
+- 列出找到的相关草稿文件（供 Execution Agent 作为本地第一手资料使用）
+
+## 步骤三：检查已有研究
+
+- 搜索 `{研究目录}/` 以避免重复
+- 如果存在相关研究：标记为更新而非新建
+
+## 步骤四：确定知识领域
+
+- 推断相关领域（如 SoftwareEngineering、Finance、Health、ProductDesign、AI、Math、Art、History）
+- 如果无法确定：标记为 `TBD`，由 Orchestrator 向用户追问
+
+## 步骤五：匹配专家人格
+
+- 列出 `{系统目录}/提示词/` 下所有 `.md` 文件
+- 逐一读取每个文件的 **frontmatter**（`id`、`aliases`、`tags`）以及 **领域覆盖** / **核心职责** 章节
+- 将研究主题（来自步骤四）与每个 persona 的领域覆盖进行比对，找出最佳匹配
+- 若多个 persona 均可适用，选择领域最具针对性的一个
+- 若没有任何 persona 的领域与研究主题重叠，标注"未找到匹配，使用通用研究模式"
+
+**适用模式判断：**
+
+- `全量应用`：研究主题完全落在该 persona 的核心领域内——用专家 Output Format 完整替换默认章节结构
+- `参考应用`：该 persona 的分析框架相关但主题并非其核心聚焦——仅借用分析框架，保留默认章节
+- `不适用`：无匹配 persona——使用默认报告结构
+
+## 步骤六：创建计划文件
+
+路径：`{计划目录}/Plan_YYYY-MM-DD_Research_Topic.md`
+
+```markdown
+---
+title: "Plan: [主题]"
+type: plan
+status: active
+created: "YYYY-MM-DD"
+source: research
+topic: "[主题]"
+tags: [plan, research]
+aliases: []
+---
+
+# 研究计划: [主题]
+
+## 触发方式
+
+[FILE MODE: {草稿目录}/文件名.md | TOPIC MODE: 主题关键词]
+
+## 研究目标
+
+[完成此研究后用户将理解什么]
+
+## 本地草稿资料
+
+[列出 {草稿目录}/ 中相关文件路径，或"未找到相关草稿"]
+
+## 现有研究
+
+[列出 {研究目录}/ 中相关的现有研究报告，或"未找到"]
+
+## 相关领域
+
+[领域名称，或 TBD]
+
+## 专家人格
+
+- 文件路径: [{系统目录}/提示词/XXX.md，或"未找到"]
+- 适用理由: [为何此 persona 适合本研究主题，一句话说明]
+- 适用模式: [全量应用 | 参考应用 | 不适用]
+- 特殊格式要求: [如 LaTeX 公式、史料引用格式等，或"无"]
+
+## 研究策略
+
+[ ] 读取本地草稿作为一手资料（如有）
+[ ] 搜索官方文档/权威资料
+[ ] 形成概念框架（术语表、模块关系、关键机制）
+[ ] 查找实际示例和用例（如适用）
+[ ] 总结最佳实践与常见陷阱
+[ ] 输出可执行的下一步建议
+
+## 输出结构（仅产出到 {研究目录}/）
+
+- 主笔记（研究报告）: {研究目录}/Domain/Topic/Topic.md
+- 示例/资源（可选）: {研究目录}/Domain/Topic/examples/
+- 可视化（可选）: {研究目录}/Domain/Topic/Topic_Map.canvas
+
+## 澄清问题回答
+
+知识水平: [待用户填写]
+方法偏好: [待用户填写]
+```
+
+## 步骤七：返回结果
+
+返回计划文件的路径。