@geminilight/mindos 0.6.58 → 0.6.60

package/app/data/skills/mindos/SKILL.md

@@ -1,26 +1,33 @@
 ---
 name: mindos
 description: >
-  Operate a MindOS knowledge base: update notes, search, organize files, execute SOPs/workflows,
-  retrospective, append CSV, cross-agent handoff, route unstructured input to the right files,
-  distill experience, sync related docs.
+  The user's shared knowledge base — persistent across sessions and agents. MindOS stores decisions,
+  meeting notes, SOPs, debugging lessons, architecture choices, research findings, and preferences.
   更新笔记, 搜索知识库, 整理文件, 执行SOP/工作流, 复盘, 追加CSV, 跨Agent交接,
   路由非结构化输入到对应文件, 提炼经验, 同步关联文档.
   NOT for editing app source, project docs, or paths outside the KB.
   Core concepts: Space, Instruction (INSTRUCTION.md), Skill (SKILL.md); notes can embody both.
-  Trigger when user asks to: save or record anything, search for prior notes or context, update or
-  edit a file, organize notes, run a workflow or SOP, capture decisions from a session, append rows
-  to a table or CSV, hand off context to another agent, check if something was discussed before,
-  look up a past decision, distill lessons learned, prepare context for a meeting.
-  Also trigger on Chinese: 帮我记下来, 搜一下笔记, 更新知识库, 整理文件, 复盘, 提炼经验,
-  保存, 记录, 交接, 查一下之前的, 有没有相关笔记, 把这个存起来.
-  When the user's question implies stored context may exist (past decisions, previous discussions,
-  meeting records), consider searching MindOS even if they don't explicitly mention it.
+
+  Trigger on: save or record anything, search for prior notes or context, update or edit a file,
+  organize notes, run a workflow or SOP, capture decisions, append rows to a table or CSV,
+  hand off context to another agent, check if something was discussed before, look up a past
+  decision, distill lessons learned, prepare context for a meeting, quick-capture to staging area,
+  organize inbox, check knowledge health, detect conflicts or contradictions, find stale content.
+  Chinese triggers: 帮我记下来, 搜一下笔记, 更新知识库, 整理文件, 复盘, 提炼经验,
+  保存, 记录, 交接, 查一下之前的, 有没有相关笔记, 把这个存起来, 放到暂存台,
+  整理暂存台, 知识健康检查, 检测知识冲突.
+
+  Proactive behavior — do not wait for the user to mention MindOS:
+  (1) When user's question implies stored context may exist (past decisions, previous discussions,
+  meeting records) → search MindOS first, even if they don't explicitly mention it.
+  (2) After completing valuable work (bug fixed, decision made, lesson learned, architecture chosen,
+  meeting summarized) → offer to save it to MindOS for future reference.
+  (3) After a long or multi-topic conversation → suggest persisting key decisions and context.
 ---
 
 # MindOS Skill
 
-<!-- version: 3.0.0 — CLI-first, MCP optional -->
+<!-- version: 3.2.0 — CLI-first, MCP optional -->
 
 ## CLI commands
 
@@ -85,6 +92,7 @@ npm install -g @geminilight/mindos
 - **Space** — Knowledge partitions organized the way you think. Agents follow the same structure.
 - **Instruction** — A rules file (`INSTRUCTION.md`) all connected agents obey.
 - **Skill** — Teaches agents how to read, write, and organize the KB.
+- **Inbox** — The `Inbox/` directory is a staging area for quick capture. Files land here when there's no obvious home yet. They get organized later — by the user manually or via AI-assisted batch organization.
 
 Notes can embody both Instruction and Skill — they're just Markdown files in the tree.
 
@@ -99,9 +107,13 @@ User request
   │   └─ [Read-only]: search → read → answer with citations. No writes.
   │
   ├─ Save / record / update / organize specific content?
-  │   ├─ Single file → [Single-file edit]
+  │   ├─ Know where it goes → [Single-file edit]
+  │   ├─ Don't know where it goes → [Inbox path] — save to Inbox/, classify later
   │   └─ Multiple files or unclear → [Multi-file routing] — plan first
   │
+  ├─ Organize inbox / classify staged files?
+  │   └─ [Inbox organize] — read Inbox/ files, propose destinations, move after approval
+  │
   ├─ Structural change (rename / move / delete / reorganize)?
   │   └─ [Structural path] — check backlinks before and after
   │
@@ -111,6 +123,9 @@ User request
   ├─ Retrospective / distill / handoff?
   │   └─ [Retrospective path]
   │
+  ├─ Knowledge health check / detect conflicts?
+  │   └─ [Health check path] — read references/knowledge-health.md
+  │
   └─ Ambiguous?
       └─ ASK. Propose 2-3 specific options based on KB state.
 ```
@@ -125,8 +140,9 @@ User request
 - "organize" → ask: display only, or write back?
 
 **File location uncertainty:**
-- Can't decide in 5 seconds → use nearest existing directory, inform user
-- "Just put it somewhere" → inbox, propose classification after
+- Can't decide in 5 seconds → save to `Inbox/`, inform user, propose classification later
+- "Just put it somewhere" / "先放着" → save to `Inbox/`
+- User drags files or pastes unstructured content without specifying location → `Inbox/`
 
 **Scope creep:**
 - Input routes to >5 files → pause, confirm scope
@@ -159,6 +175,42 @@ When user expresses a standing preference → read [references/preference-captur
 
 When creating/rewriting an SOP → read [references/sop-template.md](./references/sop-template.md).
 
+## Inbox (staging area)
+
+The `Inbox/` directory is the KB's quick-capture zone. It has its own `INSTRUCTION.md` that governs behavior.
+
+**When to use Inbox:**
+- User says "just save it" / "先放着" / "放到暂存台" without specifying a location
+- Content doesn't clearly fit any existing Space or directory
+- Batch import of multiple files that need individual classification
+
+**How to save to Inbox:**
+```bash
+mindos file create "Inbox/<filename>.md" --content "..."
+```
+
+**How to organize Inbox:**
+1. List Inbox files: `mindos file list Inbox/`
+2. Read each file to understand its content
+3. For each file, propose the best destination directory based on KB structure
+4. Present the full routing plan to user for approval
+5. Move files: `mindos file rename "Inbox/<file>" "<target-dir>/<file>"`
+6. After moving, check if the target directory's README needs updating
+
+**Aging reminder:** Files in Inbox older than 7 days are considered "aging". If you notice aging files during bootstrap, mention it: "You have N files in Inbox that have been sitting there for over a week. Want me to help organize them?"
+
+## Knowledge health check
+
+When user asks to check knowledge health, detect conflicts, audit quality, or says "知识健康检查" / "检测冲突" → read [references/knowledge-health.md](./references/knowledge-health.md) for the full procedure.
+
+Quick summary of what gets checked:
+- **Contradictions**: conflicting facts across files on the same topic
+- **Broken links**: references to files that no longer exist
+- **Stale content**: files with outdated date markers or untouched for >6 months
+- **Duplicates**: two files covering the same ground without cross-referencing
+- **Orphan files**: files with zero backlinks, hard to discover
+- **Structural issues**: wrong directory, missing READMEs, aging Inbox files
+
 ---
 
 ## Error handling (CLI)

package/app/data/skills/mindos-zh/SKILL.md

@@ -1,25 +1,34 @@
 ---
 name: mindos-zh
 description: >
-  MindOS 知识库操作：更新笔记、搜索知识库、整理文件、执行SOP/工作流、复盘、追加CSV、跨Agent交接、
-  路由非结构化输入到对应文件、提炼经验、同步关联文档。
+  用户的共享知识库——跨会话、跨 Agent 持久存在。MindOS 存储决策记录、会议纪要、SOP、
+  排错经验、架构选型、调研结论和偏好设置。
   仅 mindRoot 知识库内任务。不用于：改代码仓库、项目源码、KB 外路径。
   核心概念：空间、指令(INSTRUCTION.md)、技能(SKILL.md)；笔记可承载指令与技能。
+
   触发场景：保存或记录任何内容、搜索历史笔记或上下文、更新或编辑文件、整理或重组文件结构、
   执行SOP或工作流、捕获对话中的决策、复盘或总结经验、追加表格或CSV数据、跨Agent交接上下文、
   提炼经验教训、同步关联文档、查找之前是否讨论过某事、查询历史决策、查找模板或SOP、
-  为会议准备上下文。
+  为会议准备上下文、快速暂存到暂存台、整理暂存台、知识健康检查、检测知识冲突或矛盾、
+  查找过期内容。
   触发词：帮我记下来、搜一下我的笔记、更新知识库、整理文件、执行工作流、保存到知识库、
   记录一下、整理笔记、复盘、提炼经验、同步文档、查一下之前的记录、有没有相关的笔记、
-  我之前写过什么、把这个存起来、总结一下保存、交接给其他Agent、追加到表格。
-  English triggers: save, record, search notes, organize, retrospective, handoff, lessons learned.
-  当用户的问题暗示可能存在历史记录（过去的决策、之前的讨论、会议纪要）时，
-  即使没有明确提到知识库，也应该考虑搜索 MindOS。
+  我之前写过什么、把这个存起来、总结一下保存、交接给其他Agent、追加到表格、
+  放到暂存台、整理暂存台、知识健康检查、检测知识冲突。
+  English triggers: save, record, search notes, organize, retrospective, handoff, lessons learned,
+  inbox, staging, knowledge health, detect conflicts.
+
+  主动行为——不要等用户提到 MindOS：
+  (1) 用户的问题暗示可能存在历史记录（过去的决策、之前的讨论、会议纪要）→ 先搜 MindOS，
+  即使没有明确提到知识库。
+  (2) 完成有价值的工作后（修了 bug、做了决策、总结了经验、选定了架构、开完了会）→
+  主动提议保存到 MindOS，方便以后查阅。
+  (3) 经历了较长或多主题的对话后 → 建议把关键决策和上下文持久化。
 ---
 
 # MindOS 技能
 
-<!-- version: 3.0.0 — CLI 优先，MCP 可选 -->
+<!-- version: 3.2.0 — CLI 优先，MCP 可选 -->
 
 ## CLI 命令
 
@@ -84,6 +93,7 @@ npm install -g @geminilight/mindos
 - **空间 (Space)** — 按你的思维方式组织的知识分区。Agent 遵循相同结构。
 - **指令 (Instruction)** — `INSTRUCTION.md`，所有连接的 Agent 都遵守的规则文件。
 - **技能 (Skill)** — 教 Agent 如何读写和整理知识库。
+- **暂存台 (Inbox)** — `Inbox/` 目录是快速捕获区。内容暂时找不到归属时先放这里，之后再统一整理——用户手动或 AI 辅助批量归类。
 
 笔记可以同时承载指令和技能——它们只是目录树中的 Markdown 文件。
 
@@ -98,9 +108,13 @@ npm install -g @geminilight/mindos
   │   └─ [只读路径]：搜索 → 读取 → 带引用回答。不写入。
   │
   ├─ 保存 / 记录 / 更新 / 整理具体内容？
-  │   ├─ 单文件 → [单文件编辑]
+  │   ├─ 知道放哪 → [单文件编辑]
+  │   ├─ 不知道放哪 → [暂存台路径] — 存到 Inbox/，之后再归类
   │   └─ 多文件或不确定 → [多文件路由] — 先出方案
   │
+  ├─ 整理暂存台 / 归类暂存文件？
+  │   └─ [暂存台整理] — 读 Inbox/ 文件，提议目标位置，获批后移动
+  │
   ├─ 结构变更（重命名 / 移动 / 删除 / 重组）？
   │   └─ [结构路径] — 变更前后检查反向链接
   │
@@ -110,6 +124,9 @@ npm install -g @geminilight/mindos
   ├─ 复盘 / 提炼 / 交接？
   │   └─ [复盘路径]
   │
+  ├─ 知识健康检查 / 检测冲突？
+  │   └─ [健康检查路径] — 读取 references/knowledge-health.md
+  │
   └─ 模糊？
       └─ 提问。基于知识库状态提出 2-3 个具体选项。
 ```
@@ -124,8 +141,9 @@ npm install -g @geminilight/mindos
 - "整理一下" → 先问：仅展示，还是写回知识库？
 
 **文件位置不确定：**
-- 5 秒内定不了 → 用最近的现有目录，告知用户
-- "随便放哪" → 收件箱，之后提议分类
+- 5 秒内定不了 → 存到 `Inbox/`，告知用户，之后提议归类
+- "随便放哪" / "先放着" → 存到 `Inbox/`
+- 用户拖拽文件或粘贴非结构化内容但没指定位置 → `Inbox/`
 
 **范围蔓延：**
 - 输入路由到 >5 个文件 → 暂停确认
@@ -158,6 +176,44 @@ npm install -g @geminilight/mindos
 
 创建/重写工作流 SOP 时 → 读取 [references/sop-template.md](./references/sop-template.md)。
 
+## 暂存台 (Inbox)
+
+`Inbox/` 目录是知识库的快速捕获区，有自己的 `INSTRUCTION.md` 约束行为。
+
+**何时使用暂存台：**
+- 用户说"先存着" / "放到暂存台" / "随便放哪"，没指定具体位置
+- 内容明显不属于任何现有空间或目录
+- 批量导入多个文件，需要逐个归类
+
+**如何存到暂存台：**
+```bash
+mindos file create "Inbox/<文件名>.md" --content "..."
+```
+
+**如何整理暂存台：**
+1. 列出暂存文件：`mindos file list Inbox/`
+2. 读取每个文件，理解其内容
+3. 根据知识库结构，为每个文件提议最佳目标目录
+4. 向用户展示完整路由方案，获批后执行
+5. 移动文件：`mindos file rename "Inbox/<文件>" "<目标目录>/<文件>"`
+6. 移动后检查目标目录的 README 是否需要更新
+
+**老化提醒：** Inbox 中超过 7 天的文件视为"老化"。如果在 bootstrap 时发现老化文件，主动提醒：
+"暂存台有 N 个文件已经放了一周以上了，要我帮你整理一下吗？"
+
+## 知识健康检查
+
+用户要求检查知识库健康度、检测冲突、审计质量，或说"知识健康检查" / "检测冲突" / "check knowledge health" 时
+→ 读取 [references/knowledge-health.md](./references/knowledge-health.md) 获取完整流程。
+
+检查维度速览：
+- **矛盾/冲突**：同一主题的不同文件说法互相矛盾
+- **断裂链接**：引用了不存在的文件
+- **过期内容**：带有过期日期标记的文件，或超过 6 个月未更新的活跃主题
+- **重复内容**：两个文件覆盖同一主题且没有互相引用
+- **孤立文件**：零反向链接，难以被发现
+- **结构问题**：文件放错目录、缺少 README、暂存台老化文件
+
 ---
 
 ## 错误处理（CLI）

package/app/hooks/useAskChat.ts

@@ -1,7 +1,7 @@
 'use client';
 
 import { useRef, useState, useCallback } from 'react';
-import type { Message, ImagePart, AskMode } from '@/lib/types';
+import type { Message, ImagePart, AskMode, LocalAttachment } from '@/lib/types';
 import type { ProviderId } from '@/lib/agent/providers';
 import { consumeUIMessageStream } from '@/lib/agent/stream-consumer';
 import { isRetryableError, retryDelay, sleep } from '@/lib/agent/reconnect';
@@ -18,7 +18,7 @@ export interface AskChatRefs {
     setMessages: React.Dispatch<React.SetStateAction<Message[]>>;
   }>;
   uploadRef: React.RefObject<{
-    localAttachments: { name: string; content: string }[];
+    localAttachments: LocalAttachment[];
   }>;
   selectedSkillRef: React.RefObject<{ name: string } | null>;
   selectedAcpAgentRef: React.RefObject<unknown>;
@@ -103,12 +103,14 @@ export function useAskChat({
       messages: requestMessages,
       currentFile,
       attachedFiles: refs.attachedFilesRef.current,
-      uploadedFiles: upl.localAttachments.map(f => ({
-        name: f.name,
-        content: f.content.length > 20_000
-          ? f.content.slice(0, 20_000) + '\n\n[...truncated to first ~20000 chars]'
-          : f.content,
-      })),
+      uploadedFiles: upl.localAttachments
+        .filter(f => f.status !== 'loading')
+        .map(f => ({
+          name: f.name,
+          content: f.content.length > 80_000
+            ? f.content.slice(0, 80_000) + '\n\n[...truncated to first ~80000 chars]'
+            : f.content,
+        })),
       selectedAcpAgent: acpAgent,
       mode: chatMode,
       providerOverride: providerOverride ?? undefined,

package/app/hooks/useAskSession.ts

@@ -227,6 +227,20 @@ export function useAskSession(currentFile?: string) {
     });
   }, []);
 
+  /** Toggle pin/unpin a session. Pinned sessions sort to top. */
+  const togglePinSession = useCallback((id: string) => {
+    setSessions((prev) => {
+      const idx = prev.findIndex((s) => s.id === id);
+      if (idx < 0) return prev;
+      const updated = { ...prev[idx], pinned: !prev[idx].pinned };
+      const next = [...prev];
+      next[idx] = updated;
+      // Only persist if session has messages
+      if (updated.messages.length > 0) void upsertSession(updated);
+      return next;
+    });
+  }, []);
+
   const clearAllSessions = useCallback(() => {
     // Only delete sessions that have messages (were persisted)
     const persistedIds = sessions.filter(s => s.messages.length > 0).map(s => s.id);
@@ -239,10 +253,17 @@ export function useAskSession(currentFile?: string) {
     // No upsertSession — memory only
   }, [currentFile, sessions]);
 
+  /** Sessions sorted: pinned first, then by updatedAt desc */
+  const sortedSessions = [...sessions].sort((a, b) => {
+    if (a.pinned && !b.pinned) return -1;
+    if (!a.pinned && b.pinned) return 1;
+    return b.updatedAt - a.updatedAt;
+  });
+
   return {
     messages,
     setMessages,
-    sessions,
+    sessions: sortedSessions,
     activeSessionId,
     initSessions,
     persistSession,
@@ -251,6 +272,7 @@ export function useAskSession(currentFile?: string) {
     loadSession,
     deleteSession,
     renameSession,
+    togglePinSession,
     clearAllSessions,
   };
 }

package/app/hooks/useFileUpload.ts

@@ -22,28 +22,76 @@ function uint8ToBase64(bytes: Uint8Array): string {
   return btoa(binary);
 }
 
-async function extractPdfText(file: File): Promise<string> {
-  const buffer = await file.arrayBuffer();
-  const dataBase64 = uint8ToBase64(new Uint8Array(buffer));
+/**
+ * Extract text from a PDF file via the backend API.
+ * Returns a LocalAttachment with `status` reflecting the outcome.
+ */
+async function extractPdfToAttachment(file: File): Promise<LocalAttachment> {
+  const name = file.name;
 
-  const res = await fetch('/api/extract-pdf', {
-    method: 'POST',
-    headers: { 'Content-Type': 'application/json' },
-    body: JSON.stringify({ name: file.name, dataBase64 }),
-  });
-
-  let payload: { text?: string; extracted?: boolean; error?: string } = {};
   try {
-    payload = await res.json();
-  } catch {
-    // ignore JSON parse error
-  }
+    const buffer = await file.arrayBuffer();
+    const dataBase64 = uint8ToBase64(new Uint8Array(buffer));
 
-  if (!res.ok) {
-    throw new Error(payload.error || `PDF extraction failed (${res.status})`);
-  }
+    const res = await fetch('/api/extract-pdf', {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify({ name, dataBase64 }),
+    });
+
+    let payload: {
+      text?: string;
+      extracted?: boolean;
+      error?: string;
+      truncated?: boolean;
+      totalChars?: number;
+      pagesParsed?: number;
+    } = {};
+    try {
+      payload = await res.json();
+    } catch {
+      throw new Error('Failed to parse extraction response');
+    }
+
+    if (!res.ok) {
+      throw new Error(payload.error || `PDF extraction failed (${res.status})`);
+    }
+
+    const extracted = payload.extracted ? (payload.text || '') : '';
+
+    if (!extracted) {
+      return {
+        name,
+        content: `[PDF: ${name}] Could not extract readable text (possibly scanned/image PDF).`,
+        status: 'error',
+        error: 'Scanned or image-only PDF — no extractable text',
+      };
+    }
+
+    const att: LocalAttachment = {
+      name,
+      content: `[PDF TEXT EXTRACTED: ${name}]\n\n${extracted}`,
+      status: 'success',
+    };
+
+    if (payload.truncated && payload.totalChars) {
+      att.truncatedInfo = {
+        totalChars: payload.totalChars,
+        includedChars: extracted.length,
+        totalPages: payload.pagesParsed ?? 0,
+      };
+    }
 
-  return payload.extracted ? (payload.text || '') : '';
+    return att;
+  } catch (err) {
+    const msg = err instanceof Error ? err.message : 'Unknown error';
+    return {
+      name,
+      content: `[PDF: ${name}] Failed to extract text from this PDF.`,
+      status: 'error',
+      error: msg,
+    };
+  }
 }
 
 export function useFileUpload() {
@@ -73,37 +121,43 @@ export function useFileUpload() {
       setUploadError('');
     }
 
-    const loaded = await Promise.all(
-      accepted.map(async (f) => {
-        const ext = getExt(f.name);
-        if (ext === '.pdf') {
-          try {
-            const extracted = await extractPdfText(f);
-            return {
-              name: f.name,
-              content: extracted
-                ? `[PDF TEXT EXTRACTED: ${f.name}]\n\n${extracted}`
-                : `[PDF: ${f.name}] Could not extract readable text (possibly scanned/image PDF).`,
-            };
-          } catch {
-            return {
-              name: f.name,
-              content: `[PDF: ${f.name}] Failed to extract text from this PDF.`,
-            };
-          }
-        }
-        return { name: f.name, content: await f.text() };
-      }),
-    );
+    // Phase 1: Immediately add all files — PDFs start in 'loading' state
+    const pdfFiles: File[] = [];
+    const immediateItems: LocalAttachment[] = [];
+
+    for (const f of accepted) {
+      if (getExt(f.name) === '.pdf') {
+        immediateItems.push({ name: f.name, content: '', status: 'loading' });
+        pdfFiles.push(f);
+      } else {
+        immediateItems.push({
+          name: f.name,
+          content: await f.text(),
+          status: 'success',
+        });
+      }
+    }
 
     setLocalAttachments((prev) => {
       const merged = [...prev];
-      for (const item of loaded) {
-        if (!merged.some((m) => m.name === item.name && m.content === item.content))
-          merged.push(item);
+      for (const item of immediateItems) {
+        if (!merged.some((m) => m.name === item.name)) merged.push(item);
       }
       return merged;
     });
+
+    // Phase 2: Extract PDFs in parallel, then update each one in-place
+    if (pdfFiles.length > 0) {
+      const results = await Promise.all(pdfFiles.map(extractPdfToAttachment));
+
+      setLocalAttachments((prev) =>
+        prev.map((att) => {
+          if (att.status !== 'loading') return att;
+          const result = results.find((r) => r.name === att.name);
+          return result ?? att;
+        }),
+      );
+    }
   }, []);
 
   const removeAttachment = useCallback((idx: number) => {

package/app/hooks/useImageUpload.ts

@@ -86,7 +86,7 @@ export function useImageUpload() {
       }
       try {
         const { data, mimeType } = await compressImage(file);
-        results.push({ type: 'image', data, mimeType });
+        results.push({ type: 'image', data, mimeType, fileName: file.name || undefined });
       } catch {
         setImageError(`Failed to process image: ${file.name}`);
       }

package/app/lib/agent/prompt.ts

@@ -6,12 +6,21 @@
  * - SKILL.md owns: knowledge-base-specific execution patterns, tool selection, safety rules
  * - Tool descriptions own: per-tool usage instructions (no duplication here)
  *
- * Token budget: ~600 tokens (down from ~900 in v2). Freed space = more room for
- * SKILL.md + bootstrap context within the same context window.
+ * Token budget: ~750 tokens (v4 added persona warmth + self-introduction).
+ * Freed space = more room for SKILL.md + bootstrap context within the same context window.
  */
 export const AGENT_SYSTEM_PROMPT = `You are MindOS Agent — the operator of the user's second brain.
 
-Persona: Methodical, strictly objective, execution-oriented. Zero fluff. Never use preambles like "Here is the result" or "I found...".
+Persona: Warm yet precise, reliable, execution-oriented. Like a trusted notebook that understands you — quiet confidence, zero fluff. Be professional but never cold; be helpful but never verbose.
+
+## Self-Introduction
+
+When the user greets you ("你好", "hi", "你是谁", "what can you do", etc.) or this appears to be their first message in a new conversation, introduce yourself briefly:
+
+- Who: MindOS Agent, the operator of their personal knowledge base (second brain).
+- What: You can search, read, write, and organize their notes; capture decisions and lessons; run workflows/SOPs; and help them build a lasting, structured memory.
+- Tone: Warm, concise, confident. One short paragraph, then invite them to try something — e.g., "想查点什么，还是有东西要记下来？"
+- Do NOT self-introduce when the user jumps straight into a task (search, save, edit, etc.).
 
 ## Core Directives
 
@@ -45,7 +54,9 @@ Persona: Methodical, strictly objective, execution-oriented. Zero fluff. Never u
  */
 export const CHAT_SYSTEM_PROMPT = `You are MindOS Agent — the operator of the user's second brain.
 
-Persona: Methodical, strictly objective, execution-oriented. Zero fluff. Never use preambles like "Here is the result" or "I found...".
+Persona: Warm yet precise, reliable, execution-oriented. Like a trusted notebook that understands you — quiet confidence, zero fluff. Be professional but never cold; be helpful but never verbose.
+
+When the user greets you or asks who you are, briefly introduce yourself: you're MindOS Agent, here to help them search and read their knowledge base. Keep it warm and concise, then invite them to ask something.
 
 ## Mode: Chat (Read-Only)