dsh-plugin-t-expert 0.4.26 → 0.4.37

package/data/experts/engineering/llm-wiki/skills/llm-wiki/SKILL.md

@@ -0,0 +1,161 @@
+---
+name: llm-wiki
+description: "Build and maintain a personal knowledge base (wiki) using LLMs. Instead of RAG-style retrieval, the LLM incrementally compiles, cross-references, and maintains a persistent structured wiki from raw sources. Use when user wants to create a knowledge base, build a personal wiki, organize research notes, ingest documents into a structured wiki, or maintain a living knowledge repository."
+description_zh: "用 LLM 增量构建和维护个人知识库 Wiki"
+description_en: "Build a persistent personal wiki with LLM-maintained knowledge"
+homepage: https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f
+allowed-tools: Read,Write,Bash,Grep,Glob
+---
+
+# Karpathy LLM Wiki
+
+A pattern for building personal knowledge bases using LLMs.
+
+> Based on [Andrej Karpathy's LLM Wiki](https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f).
+
+## The Core Idea
+
+Instead of RAG (retrieve-on-every-query), the LLM **incrementally builds and maintains a persistent wiki** — a structured, interlinked collection of markdown files. When you add a new source, the LLM reads it, extracts key information, and integrates it into the existing wiki — updating entity pages, revising topic summaries, noting contradictions, and strengthening the evolving synthesis.
+
+**The wiki is a persistent, compounding artifact.** Cross-references are already there. Contradictions have been flagged. Synthesis reflects everything you've read. It gets richer with every source and every question.
+
+You never write the wiki yourself — the LLM writes and maintains all of it. You're in charge of sourcing, exploration, and asking the right questions.
+
+## Architecture (Three Layers)
+
+### 1. Raw Sources
+Your curated collection of source documents — articles, papers, images, data files. These are **immutable**. The LLM reads from them but never modifies them. This is your source of truth.
+
+### 2. The Wiki
+A directory of LLM-generated markdown files — summaries, entity pages, concept pages, comparisons, overview, synthesis. The LLM owns this layer entirely: creates pages, updates them when new sources arrive, maintains cross-references, keeps everything consistent. **You read it; the LLM writes it.**
+
+### 3. The Schema
+A configuration document (this skill) that tells the LLM the wiki's structure, conventions, and workflows. You and the LLM co-evolve this over time.
+
+## Operations
+
+### Ingest
+User drops a new source. The LLM:
+1. Reads the source and discusses key takeaways
+2. Writes a summary page in the wiki
+3. Updates the index
+4. Updates relevant entity and concept pages across the wiki
+5. Appends an entry to the log
+
+A single source may touch 10-15 wiki pages. Prefer ingesting one source at a time for quality.
+
+### Query
+Ask questions against the wiki. The LLM:
+1. Reads `index.md` to find relevant pages
+2. Reads those pages and synthesizes an answer with citations
+3. **Good answers get filed back into the wiki as new pages** — comparisons, analyses, connections should be persisted, not lost in chat history
+
+### Lint (Health Check)
+Periodically health-check the wiki. Look for:
+- Contradictions between pages
+- Stale claims superseded by newer sources
+- Orphan pages with no inbound links
+- Important concepts mentioned but lacking their own page
+- Missing cross-references
+- Data gaps that could be filled with a web search
+
+## Default Location
+
+The wiki lives at a **user-level** path so it's accessible from any workspace or conversation:
+
+```
+~/.workbuddy/wiki-knowledge/
+```
+
+This is the default. The LLM should:
+1. **Always check `~/.workbuddy/wiki-knowledge/` first** when the user mentions their wiki, regardless of the current workspace.
+2. If the directory exists and contains `WIKI-SCHEMA.md`, use it directly — no need to ask the user for a path.
+3. Only ask for a custom path if the user explicitly wants a separate wiki.
+
+## Getting Started
+
+When a user wants to create a wiki, follow these steps:
+
+### Step 1: Initialize Structure
+Create at `~/.workbuddy/wiki-knowledge/` (or a user-specified path):
+```
+wiki-knowledge/
+├── raw/              # Immutable source documents
+│   └── assets/       # Downloaded images
+├── wiki/             # LLM-maintained markdown files
+│   ├── index.md      # Content catalog
+│   └── log.md        # Chronological operation log
+└── WIKI-SCHEMA.md    # Schema & conventions (co-evolve with user)
+```
+
+### Step 2: Create index.md
+A catalog of everything in the wiki — each page listed with a link, a one-line summary, and metadata (date, source count). Organized by category. The LLM updates it on every ingest.
+
+### Step 3: Create log.md
+Append-only chronological record. Each entry format:
+```markdown
+## [YYYY-MM-DD] operation_type | Title
+Brief description of what was done.
+```
+Use consistent prefixes so it's parseable with grep:
+```bash
+grep "^## \[" log.md | tail -5
+```
+
+### Step 4: Create WIKI-SCHEMA.md
+A configuration file describing:
+- Directory structure and conventions
+- Page format templates
+- Cross-referencing rules
+- Domain-specific categories
+- Workflows for ingest/query/lint
+
+Customize this with the user for their specific domain.
+
+## Page Format Template
+
+```markdown
+---
+title: Page Title
+type: entity | concept | source-summary | comparison | synthesis
+created: YYYY-MM-DD
+updated: YYYY-MM-DD
+sources: [list of source filenames]
+tags: [relevant tags]
+---
+
+# Page Title
+
+Content here. Use [[wiki-links]] for cross-references.
+
+## See Also
+- [[related-page-1]]
+- [[related-page-2]]
+```
+
+## Use Cases
+
+| Domain | Raw Sources | Wiki Pages |
+|--------|------------|------------|
+| **Personal** | Journal entries, articles, podcast notes | Goals, health tracking, self-improvement |
+| **Research** | Papers, articles, reports | Entities, concepts, evolving thesis |
+| **Book reading** | Chapter notes | Characters, themes, plot threads |
+| **Business** | Slack threads, meeting transcripts, docs | Internal knowledge base |
+| **Competitive analysis** | Company reports, news | Competitor profiles, comparisons |
+| **Course notes** | Lecture notes, readings | Topic summaries, concept maps |
+
+## Tips
+
+- **Obsidian** works great as the wiki viewer — graph view shows connections, Dataview plugin queries frontmatter metadata
+- **The wiki is just a git repo of markdown files** — version history, branching, and collaboration for free
+- **Marp** can generate slide decks from wiki content
+- **Web Clipper** (Obsidian plugin) converts web articles to markdown for easy source ingestion
+- **Download images locally** for persistence — set a fixed attachment directory
+
+## Important Principles
+
+1. **Raw sources are immutable** — never modify source documents
+2. **The LLM owns the wiki** — users read, LLM writes and maintains
+3. **Knowledge compounds** — every ingest and query enriches the wiki
+4. **File good answers back** — don't let valuable analysis disappear into chat
+5. **Regular lint passes** — keep the wiki healthy as it grows

package/data/experts/engineering/mcp-build-expert/.codebuddy-plugin/plugin.json

@@ -0,0 +1,56 @@
+{
+  "name": "mcp-build-expert",
+  "version": "1.0.1",
+  "description": "Expert in Model Context Protocol design and implementation",
+  "agents": [
+    "./agents/mcp-build-expert.md"
+  ],
+  "expertType": "agent",
+  "agentName": "mcp-build-expert",
+  "displayName": {
+    "en": "Rex",
+    "zh": "协议通"
+  },
+  "profession": {
+    "en": "MCP Build Expert",
+    "zh": "MCP构建专家"
+  },
+  "displayDescription": {
+    "en": "Expert in Model Context Protocol design and implementation",
+    "zh": "精通Model Context Protocol设计实现"
+  },
+  "avatar": "avatars/expert.png",
+  "categoryId": "02-Engineering",
+  "defaultInitPrompt": {
+    "zh": "我们想构建MCP(Model Context Protocol)应用,需要专业的MCP开发经验,请MCP构建专家帮我们设计MCP应用。",
+    "en": "We want to build MCP (Model Context Protocol) applications and need professional MCP development experience. Please, MCP Build Expert, help us design MCP applications."
+  },
+  "tags": [
+    {
+      "zh": "MCP构建",
+      "en": "MCP Building"
+    },
+    {
+      "zh": "协议开发",
+      "en": "Protocol Dev"
+    },
+    {
+      "zh": "工具集成",
+      "en": "Tool Integration"
+    }
+  ],
+  "quickPrompts": [
+    {
+      "zh": "我们想构建MCP(Model Context Protocol)应用,需要专业的MCP开发经验,请MCP构建专家帮我们设计MCP应用。",
+      "en": "We want to build MCP (Model Context Protocol) applications and need professional MCP development experience. Please, MCP Build Expert, help us design MCP applications."
+    },
+    {
+      "zh": "设计MCP服务架构和安全方案",
+      "en": "Design MCP service architecture and security plan"
+    },
+    {
+      "zh": "搭建MCP工具测试和部署流程",
+      "en": "Build MCP tool testing and deployment pipeline"
+    }
+  ]
+}

package/data/experts/engineering/mcp-build-expert/.downloaded_at

@@ -0,0 +1 @@
+2026-09-29T03:25:00.456632+00:00

package/data/experts/engineering/mcp-build-expert/.imported-from-workbuddy

@@ -0,0 +1,7 @@
+{
+  "plugin": "mcp-build-expert",
+  "division": "engineering",
+  "categoryId": "02-Engineering",
+  "importedAt": "2026-09-29T12:42:06",
+  "source": "workbuddy expert marketplace (COS)"
+}

package/data/experts/engineering/mcp-build-expert/README.md

@@ -0,0 +1,19 @@
+# 协议通 - MCP构建专家
+
+精通Model Context Protocol设计实现
+
+## 功能特性
+
+- **MCP 协议精通**：深入理解 Model Context Protocol 的架构与规范。
+- **Server 开发**：构建 MCP Server，暴露 tools/resources/prompts。
+- **传输与集成**：熟练运用 stdio/SSE 等传输方式对接客户端。
+- **工具设计**：设计清晰的工具接口与参数 Schema。
+- **调试与测试**：定位协议交互问题并保障稳定性。
+
+## 致谢
+
+本专家在构建过程中参考并借鉴了以下开源项目，特此致谢：
+
+- [agency-agents](https://github.com/msitarzewski/agency-agents)（MIT License）
+
+相关开源许可证文本见 [`license/`](./license) 目录。

package/data/experts/engineering/mcp-build-expert/agents/mcp-build-expert.md

@@ -0,0 +1,63 @@
+---
+name: mcp-build-expert
+description: Expert Model Context Protocol developer who designs, builds, and tests MCP servers that extend AI agent capabilities with custom tools, resources, and prompts.
+color: indigo
+emoji: 🔌
+vibe: Builds the tools that make AI agents actually useful in the real world.
+---
+
+# MCP Builder Agent
+
+You are **MCP Builder**, a specialist in building Model Context Protocol servers. You create custom tools that extend AI agent capabilities — from API integrations to database access to workflow automation.
+
+## 🧠 Your Identity & Memory
+- **Role**: MCP server development specialist
+- **Personality**: Integration-minded, API-savvy, developer-experience focused
+- **Memory**: You remember MCP protocol patterns, tool design best practices, and common integration patterns
+- **Experience**: You've built MCP servers for databases, APIs, file systems, and custom business logic
+
+## 🎯 Your Core Mission
+
+Build production-quality MCP servers:
+
+1. **Tool Design** — Clear names, typed parameters, helpful descriptions
+2. **Resource Exposure** — Expose data sources agents can read
+3. **Error Handling** — Graceful failures with actionable error messages
+4. **Security** — Input validation, auth handling, rate limiting
+5. **Testing** — Unit tests for tools, integration tests for the server
+
+## 🔧 MCP Server Structure
+
+```typescript
+// TypeScript MCP server skeleton
+import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
+import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
+import { z } from "zod";
+
+const server = new McpServer({ name: "my-server", version: "1.0.0" });
+
+server.tool("search_items", { query: z.string(), limit: z.number().optional() },
+  async ({ query, limit = 10 }) => {
+    const results = await searchDatabase(query, limit);
+    return { content: [{ type: "text", text: JSON.stringify(results, null, 2) }] };
+  }
+);
+
+const transport = new StdioServerTransport();
+await server.connect(transport);
+```
+
+## 🔧 Critical Rules
+
+1. **Descriptive tool names** — `search_users` not `query1`; agents pick tools by name
+2. **Typed parameters with Zod** — Every input validated, optional params have defaults
+3. **Structured output** — Return JSON for data, markdown for human-readable content
+4. **Fail gracefully** — Return error messages, never crash the server
+5. **Stateless tools** — Each call is independent; don't rely on call order
+6. **Test with real agents** — A tool that looks right but confuses the agent is broken
+
+## 💬 Communication Style
+- Start by understanding what capability the agent needs
+- Design the tool interface before implementing
+- Provide complete, runnable MCP server code
+- Include installation and configuration instructions

package/data/experts/engineering/mcp-build-expert/license/agency-agents.LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2024-2025 msitarzewski
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/data/experts/engineering/mcp-build-expert/persona.md

@@ -0,0 +1,64 @@
+---
+name: MCP构建专家
+nameEn: MCP Build Expert
+description: 精通Model Context Protocol设计实现
+descriptionEn: MCP Build Expert
+emoji: ⚙️
+color: "#3B82F6"
+vibe: MCP构建专家
+---
+# MCP Builder Agent
+
+You are **MCP Builder**, a specialist in building Model Context Protocol servers. You create custom tools that extend AI agent capabilities — from API integrations to database access to workflow automation.
+
+## 🧠 Your Identity & Memory
+- **Role**: MCP server development specialist
+- **Personality**: Integration-minded, API-savvy, developer-experience focused
+- **Memory**: You remember MCP protocol patterns, tool design best practices, and common integration patterns
+- **Experience**: You've built MCP servers for databases, APIs, file systems, and custom business logic
+
+## 🎯 Your Core Mission
+
+Build production-quality MCP servers:
+
+1. **Tool Design** — Clear names, typed parameters, helpful descriptions
+2. **Resource Exposure** — Expose data sources agents can read
+3. **Error Handling** — Graceful failures with actionable error messages
+4. **Security** — Input validation, auth handling, rate limiting
+5. **Testing** — Unit tests for tools, integration tests for the server
+
+## 🔧 MCP Server Structure
+
+```typescript
+// TypeScript MCP server skeleton
+import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
+import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
+import { z } from "zod";
+
+const server = new McpServer({ name: "my-server", version: "1.0.0" });
+
+server.tool("search_items", { query: z.string(), limit: z.number().optional() },
+  async ({ query, limit = 10 }) => {
+    const results = await searchDatabase(query, limit);
+    return { content: [{ type: "text", text: JSON.stringify(results, null, 2) }] };
+  }
+);
+
+const transport = new StdioServerTransport();
+await server.connect(transport);
+```
+
+## 🔧 Critical Rules
+
+1. **Descriptive tool names** — `search_users` not `query1`; agents pick tools by name
+2. **Typed parameters with Zod** — Every input validated, optional params have defaults
+3. **Structured output** — Return JSON for data, markdown for human-readable content
+4. **Fail gracefully** — Return error messages, never crash the server
+5. **Stateless tools** — Each call is independent; don't rely on call order
+6. **Test with real agents** — A tool that looks right but confuses the agent is broken
+
+## 💬 Communication Style
+- Start by understanding what capability the agent needs
+- Design the tool interface before implementing
+- Provide complete, runnable MCP server code
+- Include installation and configuration instructions

package/data/experts/engineering/mibao-media-archivist/.codebuddy-plugin/plugin.json

@@ -0,0 +1,77 @@
+{
+  "name": "mibao-media-archivist",
+  "version": "26.9.2",
+  "description": "Local-first media inventory expert for read-only scanning, strict SHA-256, exact duplicate detection, local indexing, search, and openable reports.",
+  "author": {
+    "email": "unique@u3w.com",
+    "name": "MIBAO Team"
+  },
+  "license": "Proprietary",
+  "keywords": [
+    "media",
+    "photo",
+    "video",
+    "inventory",
+    "hash",
+    "duplicate",
+    "local-search"
+  ],
+  "expertType": "agent",
+  "agentName": "mibao-media-archivist",
+  "agents": [
+    "./agents/mibao-media-archivist.md"
+  ],
+  "skills": [
+    "./skills/mibao",
+    "./skills/mibao-index-policy",
+    "./skills/mibao-safety-qa"
+  ],
+  "displayName": {
+    "en": "MIBAO",
+    "zh": "秘宝"
+  },
+  "profession": {
+    "en": "AI Media Asset Archivist",
+    "zh": "AI影像资产整理专家"
+  },
+  "displayDescription": {
+    "en": "Protects originals while building strict SHA-256 inventories, exact duplicate groups, local search indexes, and verifiable reports.",
+    "zh": "在不改动原件的前提下，完成影像盘点、严格SHA-256、精确重复识别、本地索引检索与可验收报告"
+  },
+  "avatar": "avatars/mibao.jpg",
+  "categoryId": "04-DataAI",
+  "defaultInitPrompt": {
+    "en": "Scan the media folder I authorize, compute strict SHA-256 hashes, and generate exact duplicate groups, a local search index, and an openable inventory report.",
+    "zh": "扫描我授权的影像文件夹，执行严格 SHA-256，生成精确重复项清单、本地检索索引和可打开的盘点报告"
+  },
+  "tags": [
+    {
+      "en": "Media Inventory",
+      "zh": "影像盘点"
+    },
+    {
+      "en": "Exact Duplicate Detection",
+      "zh": "精确重复识别"
+    },
+    {
+      "en": "Local Index Search",
+      "zh": "本地索引检索"
+    }
+  ],
+  "quickPrompts": [
+    {
+      "en": "Scan the media folder I authorize, compute strict SHA-256 hashes, and generate exact duplicate groups, a local search index, and an openable inventory report.",
+      "zh": "扫描我授权的影像文件夹，执行严格 SHA-256，生成精确重复项清单、本地检索索引和可打开的盘点报告"
+    },
+    {
+      "en": "Search an existing MIBAO project for my keywords and return relative paths, stable asset IDs, evidence references, and match summaries.",
+      "zh": "在已有秘宝项目中检索我指定的关键词。返回相对路径、稳定资产ID、证据引用和命中摘要"
+    },
+    {
+      "en": "Update an existing MIBAO project for new or changed files, refresh strict hashes, duplicate groups, the local index, and reports, without changing originals.",
+      "zh": "继续盘点我已有的秘宝项目，更新新增或变化文件的严格哈希、重复项、索引和报告，原件不动"
+    }
+  ],
+  "defaultEnabled": true,
+  "plugin": "mibao-media-archivist"
+}

package/data/experts/engineering/mibao-media-archivist/.downloaded_at

@@ -0,0 +1 @@
+2026-09-29T03:25:14.630580+00:00

package/data/experts/engineering/mibao-media-archivist/.imported-from-workbuddy

@@ -0,0 +1,7 @@
+{
+  "plugin": "mibao-media-archivist",
+  "division": "engineering",
+  "categoryId": "04-DataAI",
+  "importedAt": "2026-09-29T12:42:06",
+  "source": "workbuddy expert marketplace (COS)"
+}

package/data/experts/engineering/mibao-media-archivist/LICENSE

@@ -0,0 +1,9 @@
+MIBAO Proprietary License
+
+Copyright (c) 2026 MIBAO Team. All rights reserved.
+
+No permission is granted to copy, modify, distribute, sublicense, or sell this software or
+its associated assets except under a separate written agreement from the copyright holder.
+
+Third-party components, when introduced, remain subject to their own licenses and must be
+listed in THIRD_PARTY_NOTICES.md and the project SBOM.

package/data/experts/engineering/mibao-media-archivist/PRIVACY.md

@@ -0,0 +1,46 @@
+# 秘宝 26.9.2 隐私说明
+
+本说明适用于 `mibao-media-archivist` 26.9.2 的完整插件候选 ZIP 与 Agent 专家候选 ZIP。当前产品范围是 `execution_core_v1`，只包含 `inventory_build` 和 `inventory_search`。本说明描述候选包实际处理的数据与边界，不代表候选已通过活动宿主、企业上传、正式提审或上架。
+
+## 两个分发通道
+
+- Agent 专家候选不声明、安装或连接本地 MCP、连接器、localhost 服务或常驻进程。
+- 完整插件候选保留一个独立研发用 stdio MCP `mibao_ping`；它只返回非敏感运行时状态，不读取媒体。完整插件的 MCP 证据不能回填为 Agent 专家能力。
+- 两个候选都需要宿主提供可解析的 Python/SQLite 运行条件；运行时不在 ZIP 内。当前包不是自包含 Windows 运行时证明。
+
+## 读取的数据
+
+`inventory_build` 在用户明确确认的一个源目录内：
+
+- 读取绝对源路径、相对文件名、文件大小、修改时间和文件系统身份；
+- 有界读取文件头用于媒体类型识别；
+- 读取文件全部字节以执行严格 SHA-256；
+- 不移动、改名、覆盖、写回或删除源文件；
+- 当前范围不解码媒体、不做人脸识别，不执行 OCR、ASR、视觉向量或内容语义识别。
+
+`inventory_search` 只接受 `local-private` 项目，并在同一个已经审计且只读的 SQLite 快照事务上完成隐私门与 FTS 查询；结果显式返回 `privacyMode=local-private`。它不读取媒体文件，也不修改项目数据库。旧 `hybrid` 或 `deep-understanding` 项目会失败关闭；本版不提供原地隐私迁移，用户需新建或重建 `local-private` 项目后重新发起请求，也不会被静默扩大权限。
+
+## 本地写入与保留
+
+`inventory_build` 会在用户确认的项目目录持久化 `project.json`、SQLite 数据库、锁/恢复状态、FTS 索引和 HTML/CSV/JSONL/manifest 报告。数据库可包含绝对源根、相对文件名、大小、时间、文件身份、SHA-256、稳定 ID、重复关系和操作状态；报告包含相对路径、哈希、统计与异常。原始媒体不会被复制进项目。
+
+WorkBuddy 会先把包含绝对源路径、项目路径、项目名或查询词的请求 JSON 写入当前会话 Workspace Folder 下 `.workbuddy/mibao/expert-request.json`。一次性入口要求宿主同时传入该绝对请求根与 `session-workspace` 固定作用域。包内入口由 WorkBuddy 在加载 `mibao` Skill 时解析的 `${CODEBUDDY_SKILL_DIR}` 固定定位，不读取环境变量或扫描 plugin root。对严格 schema-valid、无链接/重解析点且只有一个物理链接的请求，入口会在访问源目录或项目目录前将它原子 claim、复核同一内容并删除；即使后续项目操作失败，该有效请求也已消费。错误结果以 `requestDisposition` 和 `recovery` 区分未消费、已消费和可能残留。
+
+无效请求、相对路径、错根路径、legacy 多硬链接或其他未通过所有权/完整性门的文件不保证自动清理；清理门失败时也可能残留内部 claim 文件。删除固定请求路径不等于介质安全擦除，也不能删除操作系统备份、同步副本或宿主日志。当前候选不自动删除项目数据，也尚未证明升级或卸载后的项目保留语义。
+
+## 网络、模型与外发边界
+
+秘宝候选 one-shot 子进程不包含网络客户端调用，不创建 `ExternalDisclosure`，不自动上传媒体、索引、报告、遥测或诊断数据。结果中的 `networkCallCount=0` 只描述候选 one-shot 子进程，不等同于 WorkBuddy 模型会话零网络，也不等同于整台电脑零网络。
+
+用户提示、绝对路径、项目名、查询词、工具结果、相对文件名、统计或报告回读内容可能出现在 WorkBuddy 会话中，并由当前宿主/模型按其自身服务与隐私条款处理。秘宝当前执行范围禁止把原始图像或视频作为模型证据外发，但不能把候选子进程的零网络声明外推到 WorkBuddy 聊天传输、宿主工具、`present_files`、宿主遥测或系统管理的 session-local daily memory。专家不得主动创建长期 memory、Skill 或跨项目记录。
+
+## 遥测与诊断
+
+候选包没有秘宝自有的自动遥测、崩溃上传、广告跟踪或后台更新客户端。诊断包的自动上传尚未实现。WorkBuddy 宿主自身的网络、遥测、日志和模型处理不属于候选 one-shot 子进程，需按宿主条款和独立回执判断。
+
+## 用户控制与未完成门禁
+
+- 用户在执行前确认精确源目录、项目目录和只读/写入边界；秘宝不得自动扩大扫描范围。
+- V1 不提供永久删除或覆盖原件接口。
+- 完整的旧项目隐私迁移、ExternalDisclosure 账本、媒体脱敏、逐次模型外发许可、无效/清理失败请求的处置、升级/卸载数据保留和干净 Windows 验收仍是后续门禁。
+- 本说明是工程隐私披露，不是法律意见。第一方商业分发、正式提审与上架仍需独立 owner 授权和平台回执。

package/data/experts/engineering/mibao-media-archivist/README.md

@@ -0,0 +1,37 @@
+# 秘宝 26.9.2 Agent 专家候选
+
+秘宝在不改动原件的前提下，实际完成影像盘点、严格 SHA-256、精确重复识别、SQLite FTS5 本地索引、项目检索和可打开报告。当前只开放 `inventory_build` 与 `inventory_search`，不安装或连接本地 MCP、连接器或常驻服务。
+
+## 第一次使用
+
+1. 召唤“秘宝”，选择“扫描我授权的影像文件夹…”推荐问题。
+2. 提供并确认一个绝对源目录、一个互不包含的绝对项目目录和项目显示名。
+3. 秘宝只读源目录，在项目目录生成 `project.json`、SQLite 数据库、索引和 HTML/CSV/JSONL/manifest 报告。
+4. 秘宝回读报告后给出发现文件、资产、严格重哈希、精确重复组、索引与异常数量。
+5. 后续可使用“检索已有秘宝项目”或“更新已有秘宝项目”推荐问题。
+
+严格有效的 one-shot 请求绑定宿主给出的绝对请求根与固定作用域，并在访问源目录或项目目录前被消费；无效、错根或清理失败请求不作自动清理承诺。项目检索只接受 `local-private`，旧隐私模式会失败关闭且不会自动迁移。
+
+单个不符合闭合路径策略的文件名或 reparse 条目会被明确排除，不会读取、跟随、哈希或索引；其受限相对路径和原因随 `partial` 结果返回。真正不可读、源离线、身份变化或预算超限仍阻断。项目目录可以不存在，也可以是预先创建的完全空目录。
+
+检索文件名时，点作为白名单 token 分隔符，`normal.jpg` 会同时匹配 `normal` 与 `jpg`；FTS 操作符、通配符、引号和路径符仍拒绝。
+
+重复组统计中，`content.exactDuplicateGroupCount` 表示本次严格哈希确认值；报告的严格 token 时点值是 `strict_as_of_exact_duplicate_group_count`。`current_exact_duplicate_group_count=0` 是防止把 token 后文件系统竞态误报为实时重复结论的保守语义。
+
+## 当前不包含
+
+格式转换、OCR、ASR、人物/事件识别、视觉/语义检索、片段导出、素材包和成片都未进入本版执行面。缺少报告回读、证据引用或边界回执时，不得把任务报告为完成。
+
+## 版本变更
+
+- 26.9.2（QA 修复）：修复 Windows 上含可执行文件（`.exe`/`.bat`/`.cmd`/`.com`）的源目录必然盘点失败的问题。CPython 在 Windows 上 `os.stat()` 会按扩展名合成执行位而 `os.fstat()` 不会，媒体探测的 stat↔fstat 身份比对使用了原始 `st_mode`，导致可执行文件被确定性误判为"身份已变"（MB-FS-0006 → scan_candidate_stale → fail-closed）。现改为只比较文件类型位（`stat.S_IFMT`），与哈希管线 `_same_open_identity` 的既有正确写法一致；防篡改语义（device/inode/size/mtime）不变。
+- 26.8.30：首次提交范围（`execution_core_v1`）。
+
+## 随包说明
+
+- [当前提交范围](TASKBOOK.md)
+- [隐私说明](PRIVACY.md)
+- [第一方许可](LICENSE)
+- [第三方通知](THIRD_PARTY_NOTICES.md)
+
+本包已通过本地、独立解压和 isolated Expert Manager 门；当前后继尚未通过活动 WorkBuddy 会话、自包含 Windows、企业上传、正式提审或上架门。

package/data/experts/engineering/mibao-media-archivist/TASKBOOK.md

@@ -0,0 +1,33 @@
+# 秘宝 26.9.2 专家提交范围
+
+本文件是专家 ZIP 内的提交范围投影。仓库中的完整任务书、历史研究和后续版本候选不属于本包的能力声明。
+
+## 当前能力
+
+26.9.2 只提供 `execution_core_v1`：
+
+- `inventory_build`：在用户授权目录上创建或打开本地私密项目，执行只读增量扫描、媒体类型识别、严格 SHA-256、精确重复识别、SQLite FTS5 关键词索引，并生成可回读盘点报告；
+- `inventory_search`：只在既有项目内执行 SQLite FTS5 关键词检索，返回相对路径、稳定资产标识和证据引用。
+
+盘点允许安全排除单个路径策略违规文件名或 reparse 条目并返回 `partial + scan.issues`；这些条目不读取、不跟随、不哈希、不入索引。不可读、离线、身份漂移和预算失败仍阻断。已有完全空项目目录可初始化。文件名查询中的点只作白名单 token 分隔符。
+
+两项操作均由包内固定 one-shot 入口执行。请求通过严格 JSON Schema 传入，不把用户输入拼接到 Shell 参数；请求路径必须精确绑定当前会话 Workspace Folder 下的绝对请求根和 `session-workspace` 固定作用域。包内脚本只由 WorkBuddy 加载 Skill 时解析的 `${CODEBUDDY_SKILL_DIR}` 固定定位。严格有效、单物理链接且无重解析点的请求会在源/项目访问前被原子 claim、复核并删除；无效或清理失败请求不作自动清理承诺。执行结束后退出，不安装本地 MCP、连接器或常驻服务，也不修改 WorkBuddy 配置。
+
+请求只允许当前 WorkBuddy 会话 Workspace Folder 下的 session-owned fallback，不运行 Shell/PowerShell 环境发现、目录遍历、Skill 列表、源码扫描或版本探针。每项 operation 只允许一次 request write 和一次固定 native Bash one-shot；禁止其他 Shell、`-c`、管道、重定向和拼接。报告只读回，宿主强制 memory 必须单独披露且不构成秘宝 operation。
+
+## 本版本不包含
+
+- 本地大模型、视觉向量或混合向量检索；
+- 模型状态、硬件选型、下载、安装、修复、卸载或自检入口；
+- FFmpeg、PyAV、模型权重、wheel、虚拟环境或缓存载荷；
+- 转换、OCR、ASR、片段导出、素材包或成片的产品承诺。
+
+仓库中保留的后续版本研究代码与失败回执不进入本专家 ZIP，也不构成本版本能力。
+
+## 安全与证据边界
+
+- 原件默认只读；源目录与项目目录不得重叠；
+- `inventory_search` 只接受 `local-private`，在同一已审计只读连接上查询并返回该模式；旧隐私模式失败关闭且不自动迁移；
+- 成功必须有真实项目、索引和报告回读，不能用方案或说明代替执行；
+- 当前包仍是候选。源码测试、独立解压或临时宿主审计不能证明活动安装、前台可见、真实专家会话、企业上传接受、可提审、已上架或自然使用；
+- 用户无需另装 Python 是发布目标，但须由干净 Windows 与 WorkBuddy 生命周期回执证明；未通过前保持未证明。

package/data/experts/engineering/mibao-media-archivist/THIRD_PARTY_NOTICES.md

@@ -0,0 +1,12 @@
+# MIBAO 26.9.2 package notices
+
+The WorkBuddy full-plugin and Agent-expert ZIPs contain MIBAO first-party source, policy, schema, documentation, and the project-owned expert avatar only. They bundle no third-party executable, DLL, Python wheel, virtual environment, model weight, model manifest, font, sample media, or web asset.
+
+The expert avatar was generated for this repository and contains no third-party logo, text, or watermark. MIBAO first-party material is governed by the packaged `LICENSE`.
+
+WorkBuddy, Microsoft Windows, the WorkBuddy-provided CPython runtime, and SQLite/FTS5 are external host requirements. They are not copied into, redistributed by, or licensed through either MIBAO ZIP. The separate Python engineering wheel is also outside the WorkBuddy package denominator.
+
+Future FFmpeg, metadata-tool, OCR, ASR, visual, or model candidates are not approved or bundled by this notice. Any future payload requires a new package-exact SBOM, source and binary checksums, license files, transitive notices, and owner approval.
+
+This notice is an engineering compliance statement, not legal advice. Commercial distribution, submission, and listing require separate first-party owner authorization.
+