@bgicli/bgicli 2.4.3 → 2.4.5

package/README.md

@@ -2,8 +2,8 @@
 
 **BGI CLI** 是面向中国生物学研究者的 AI 终端工具，开箱即用，无需额外配置。
 
-- ✅ **开箱即用** — `npm install -g @bgicli/bgicli` 即可
-- ✅ **内置 889 个技能** — 21 个生信工作流 + 868 个 OpenClaw 医学技能，自动安装
+- ✅ **开箱即用** — clone 仓库后一键安装即可
+- ✅ **内置 1001 个技能** — 涵盖生信分析、结构生物学、药物发现、临床等全领域，自动安装
 - ✅ **智能技能路由** — 描述任务自动激活对应技能，无需手动搜索
 - ✅ **中国 AI 服务商** — 百炼(DashScope)聚合：Qwen、DeepSeek、Kimi、MiniMax 等 20+ 模型
 - ✅ **真实工具调用** — 执行 bash、读写文件、运行 R/Python 脚本
@@ -13,26 +13,43 @@
 
 ## 安装
 
+**环境要求：** Node.js 18+
+
 ```bash
-# 需要 Node.js 18+
-npm install -g @bgicli/bgicli --registry https://registry.npmjs.org
+git clone https://gitlab.genomics.cn/ai/bgi-cli.git
+cd bgi-cli
+npm install
+npm run build
+sudo npm link
 ```
 
-首次运行自动初始化工作流和技能库（约 16MB），无需额外操作。
+> **macOS / Linux 无 root 权限时**，`npm link` 可能报 EACCES 权限错误，需先配置 npm 用户目录：
+> ```bash
+> mkdir -p ~/.npm-global
+> npm config set prefix '~/.npm-global'
+> echo 'export PATH=~/.npm-global/bin:$PATH' >> ~/.bashrc  # zsh 用户改为 ~/.zshrc
+> source ~/.bashrc
+> npm link   # 再次执行
+> ```
+
+安装完成后直接运行：
 
 ```bash
 bgi
 ```
 
+首次运行自动初始化技能库（约 16MB），无需额外操作。
+
 ---
 
 ## 卸载
 
 ```bash
-# 卸载 npm 包
-npm uninstall -g @bgicli/bgicli
+# 卸载命令行工具
+cd bgi-cli
+npm unlink
 
-# 删除本地数据（配置、工作流、技能库）
+# 删除本地数据（配置、技能库）
 # Linux / macOS
 rm -rf ~/.bgicli
 
@@ -48,7 +65,7 @@ Remove-Item -Recurse -Force "$env:USERPROFILE\.bgicli"
 bgi                    # 启动
 /connect               # 首次配置 API Key
 /cat                   # 浏览技能分类目录
-/sk deseq2             # 搜索并激活 DESeq2 工作流
+/sk deseq2             # 搜索并激活 DESeq2 技能
 /help                  # 查看全部命令
 ```
 
@@ -65,11 +82,16 @@ bgi                    # 启动
 | 模型 | 命令 |
 |------|------|
 | Qwen3.5-plus（默认） | `/model qwen3.5-plus` |
+| Qwen3.5-397B（旗舰） | `/model qwen3.5-397b-a17b` |
 | Qwen3-235B | `/model qwen3-235b-a22b` |
-| DeepSeek-R1 | `/model deepseek-r1` |
+| Qwen3-Coder-Plus | `/model qwen3-coder-plus` |
+| DeepSeek-R1（推理） | `/model deepseek-r1` |
 | DeepSeek-V3 | `/model deepseek-v3` |
+| DeepSeek-V3.2 | `/model deepseek-v3.2` |
 | Kimi-K2.5 | `/model kimi-k2.5` |
+| Kimi-K2-Thinking（推理） | `/model kimi-k2-thinking` |
 | MiniMax-M2.5 | `/model MiniMax-M2.5` |
+| GLM-5 | `/model glm-5` |
 | QwQ-Plus（推理） | `/model qwq-plus` |
 
 获取 API Key：[bailian.console.aliyun.com](https://bailian.console.aliyun.com/)
@@ -98,13 +120,15 @@ bgi                    # 启动
 | `/connect [provider]` | 配置 API Key |
 | `/status` | 显示当前配置 |
 
-### 技能与工作流
+### 技能
 | 命令 | 说明 |
 |------|------|
 | `/cat` | 按领域浏览技能分类目录（11个领域） |
-| `/sk` | 列出全部技能（工作流 + OpenClaw Medical） |
+| `/sk` | 列出全部技能 |
 | `/sk <关键词>` | 搜索并激活技能（如 `/sk deseq2`、`/sk alphafold`） |
 | `/wf` | 同 `/sk`，别名 |
+| `/install <url\|slug>` | 从 GitHub 或 SkillHub 安装新技能 |
+| `/uninstall <id>` | 卸载已安装的技能 |
 
 > **智能路由**：直接描述任务，BGI CLI 自动识别并激活对应技能。
 
@@ -176,13 +200,13 @@ bgi                    # 启动
 
 ---
 
-## 内置技能库
+## 内置技能库（1001个）
 
-### 生物信息学工作流（21个）
+使用 `/cat` 浏览分类目录，`/sk <关键词>` 搜索，`/install` 安装更多技能。
 
-#### 转录组学
-| ID | 说明 |
-|----|------|
+### 🧬 转录组学
+| 技能 ID | 说明 |
+|---------|------|
 | `bulk-rnaseq-counts-to-de-deseq2` | DESeq2 差异表达分析 |
 | `bulk-omics-clustering` | 样本/特征聚类（K-Means / HDBSCAN） |
 | `scrnaseq-scanpy-core-analysis` | 单细胞 RNA-seq（Scanpy/Python） |
@@ -192,46 +216,35 @@ bgi                    # 启动
 | `functional-enrichment-from-degs` | 功能富集（GO / KEGG / GSEA） |
 | `grn-pyscenic` | 基因调控网络（pySCENIC） |
 
-#### 基因组学
-| ID | 说明 |
-|----|------|
+### 🧪 基因组学
+| 技能 ID | 说明 |
+|---------|------|
 | `genetic-variant-annotation` | VCF 变异注释（VEP / ANNOVAR） |
 | `gwas-to-function-twas` | GWAS → TWAS 因果基因 |
 | `mendelian-randomization-twosamplemr` | 孟德尔随机化 |
 | `polygenic-risk-score-prs-catalog` | 多基因风险评分（PRS） |
 | `pooled-crispr-screens` | CRISPR 文库筛选（MAGeCK / BAGEL2） |
 
-#### 表观基因组
-| ID | 说明 |
-|----|------|
+### 🔗 表观基因组
+| 技能 ID | 说明 |
+|---------|------|
 | `chip-atlas-peak-enrichment` | ChIP-seq 峰值富集 |
 | `chip-atlas-diff-analysis` | 差异结合分析 |
 | `chip-atlas-target-genes` | 转录因子靶基因鉴定 |
 
-#### 临床与流行病学
-| ID | 说明 |
-|----|------|
+### 🏥 临床与流行病学
+| 技能 ID | 说明 |
+|---------|------|
+| `survival-analysis-clinical` | 临床生存分析（KM 曲线 / Cox 回归 / 竞争风险） |
 | `clinicaltrials-landscape` | 临床试验格局分析 |
 | `literature-preclinical` | 临床前文献系统提取 |
 | `experimental-design-statistics` | 实验设计与统计检验 |
 | `lasso-biomarker-panel` | LASSO 生物标志物筛选 |
 | `pcr-primer-design` | PCR/qPCR 引物设计 |
 
-### OpenClaw Medical Skills（868个）
-
-覆盖结构生物学、单细胞、药物发现、抗体设计、文献检索等领域，使用 `/cat` 浏览分类目录。
+### 📦 更多技能（979个）
 
----
-
-## 从源码安装
-
-```bash
-git clone https://github.com/zja2004/BGI-CLI.git
-cd BGI-CLI
-npm install
-npm run build
-npm link
-```
+结构生物学、单细胞、药物发现、抗体设计、文献检索、临床 AI 等，使用 `/cat` 或 `/sk <关键词>` 查找。
 
 ---
 
@@ -248,7 +261,7 @@ bgi
 ├── src/prompt.ts       — 生物信息学系统提示
 ├── src/providers.ts    — 中国 AI 服务商配置
 ├── src/config.ts       — 配置管理（~/.bgicli/config.json）
-└── data/               — 内置数据（工作流 + Skills + Python 工具）
+└── data/               — 内置数据（Skills + Python 工具）
 ```
 
 ---

package/dist/bgi.js

@@ -13868,7 +13868,7 @@ var TOOL_DEFINITIONS = [
     type: "function",
     function: {
       name: "read_file",
-      description: "Read the contents of a file. Use to inspect data files, SKILL.md workflow guides, R/Python scripts, results files, etc.",
+      description: "Read the contents of a file. Use to inspect data files, SKILL.md guides, R/Python scripts, results files, etc.",
       parameters: {
         type: "object",
         properties: {
@@ -13919,7 +13919,7 @@ var TOOL_DEFINITIONS = [
     type: "function",
     function: {
       name: "search_files",
-      description: "Search for files matching a pattern using glob syntax. Use to find scripts, data files, or workflow references.",
+      description: "Search for files matching a pattern using glob syntax. Use to find scripts, data files, or skill references.",
       parameters: {
         type: "object",
         properties: {
@@ -14561,7 +14561,7 @@ async function compactMessages(messages, config) {
     messages: [
       {
         role: "system",
-        content: "\u4F60\u662F\u4E00\u4E2A\u5BF9\u8BDD\u6458\u8981\u52A9\u624B\u3002\u8BF7\u5C06\u4EE5\u4E0B\u5BF9\u8BDD\u5386\u53F2\u538B\u7F29\u4E3A\u7B80\u6D01\u7684\u4E2D\u6587\u6458\u8981\uFF0C\u4FDD\u7559\u6240\u6709\u5173\u952E\u6280\u672F\u4FE1\u606F\uFF1A\u6587\u4EF6\u8DEF\u5F84\u3001\u547D\u4EE4\u3001\u5206\u6790\u7ED3\u679C\u3001\u7528\u6237\u51B3\u7B56\u3001\u5DF2\u6FC0\u6D3B\u7684\u5DE5\u4F5C\u6D41/\u6280\u80FD\u3002\u6458\u8981\u5E94\u8BA9\u5BF9\u8BDD\u80FD\u591F\u65E0\u7F1D\u7EE7\u7EED\u3002"
+        content: "\u4F60\u662F\u4E00\u4E2A\u5BF9\u8BDD\u6458\u8981\u52A9\u624B\u3002\u8BF7\u5C06\u4EE5\u4E0B\u5BF9\u8BDD\u5386\u53F2\u538B\u7F29\u4E3A\u7B80\u6D01\u7684\u4E2D\u6587\u6458\u8981\uFF0C\u4FDD\u7559\u6240\u6709\u5173\u952E\u6280\u672F\u4FE1\u606F\uFF1A\u6587\u4EF6\u8DEF\u5F84\u3001\u547D\u4EE4\u3001\u5206\u6790\u7ED3\u679C\u3001\u7528\u6237\u51B3\u7B56\u3001\u5DF2\u6FC0\u6D3B\u7684\u6280\u80FD\u3002\u6458\u8981\u5E94\u8BA9\u5BF9\u8BDD\u80FD\u591F\u65E0\u7F1D\u7EE7\u7EED\u3002"
       },
       {
         role: "user",
@@ -14603,7 +14603,7 @@ function buildSystemPrompt(dbSection) {
 ## Core Identity
 - You are purpose-built for bioinformatics, NOT a general coding assistant
 - You default to speaking Chinese unless the user writes in English
-- You always prefer using existing, validated workflow scripts over writing new code from scratch
+- You always prefer using existing, validated skill scripts over writing new code from scratch
 - You are pragmatic: suggest concrete commands, not theory
 
 ---
@@ -14611,26 +14611,53 @@ function buildSystemPrompt(dbSection) {
 ## Tool Use Policy
 You have access to these tools:
 - **bash**: Execute shell commands (R, Python, bash scripts, bioinformatics tools)
-- **read_file**: Read any file (scripts, data files, SKILL.md workflow guides)
+- **read_file**: Read any file (scripts, data files, SKILL.md guides)
 - **write_file**: Create or update files
 - **list_dir**: List directory contents
 - **search_files**: Find files by pattern (glob)
 - **fetch_geo**: Query NCBI GEO database by accession (GSE/GDS/GPL/GSM). Returns metadata, sample info, organism, platform, and ready-to-use R/Python download code. **Always call this first when the user mentions a GEO accession number \u2014 never ask them to download manually.**
 
-**MANDATORY WORKFLOW**: When the user gives you a bioinformatics task:
-1. Check if a matching pre-built workflow exists (see Workflow Library below)
-2. If yes: read the workflow's SKILL.md first, then follow it strictly
-3. If no: plan a principled approach and explain your reasoning
+## \u26A1 SKILL-FIRST PROTOCOL\uFF08\u5F3A\u5236\u6267\u884C\uFF0C\u4E0D\u53EF\u8DF3\u8FC7\uFF09
+
+\u5F53\u7528\u6237\u63D0\u51FA\u4EFB\u4F55\u751F\u7269\u4FE1\u606F\u5B66 / \u836F\u7269\u53D1\u73B0 / \u4E34\u5E8A\u5206\u6790\u4EFB\u52A1\u65F6\uFF0C**\u5FC5\u987B\u4E25\u683C\u6309\u4EE5\u4E0B\u987A\u5E8F\u6267\u884C\uFF0C\u7981\u6B62\u4E71\u5E8F**\uFF1A
+
+**\u7B2C\u4E00\u6B65 \u2014 \u8BC6\u522B\u6280\u80FD\uFF08\u5728\u8F93\u51FA\u4EFB\u4F55\u6B63\u6587\u524D\uFF09**
+\u626B\u63CF\u4E0B\u65B9 Skill Library\uFF0C\u627E\u51FA\u6240\u6709\u4E0E\u7528\u6237\u4EFB\u52A1\u76F8\u5173\u7684\u6280\u80FD\u3002\u5982\u679C Skill Library \u6CA1\u6709\u7CBE\u786E\u5339\u914D\uFF0C\u4E5F\u5E94\u641C\u7D22 Extended Skills \u76EE\u5F55\u3002
+
+**\u7B2C\u4E8C\u6B65 \u2014 \u5C55\u793A\u5019\u9009\u6280\u80FD\u5E76\u8BF7\u7528\u6237\u786E\u8BA4**
+\u7528\u5982\u4E0B\u683C\u5F0F\uFF08\u4E2D\u6587\uFF09\u544A\u77E5\u7528\u6237\uFF0C**\u4E0D\u8981\u5148\u505A\u5176\u4ED6\u4E8B**\uFF1A
+> \u{1F50D} \u68C0\u6D4B\u5230\u4EE5\u4E0B\u76F8\u5173\u6280\u80FD\uFF1A
+>
+> \u2022 \`skill-id-1\` \u2014 **[\u6280\u80FD\u540D\u79F0]**\uFF1A[\u4E00\u53E5\u8BDD\u8BF4\u660E\u4E3A\u4F55\u5339\u914D\u8BE5\u4EFB\u52A1]
+> \u2022 \`skill-id-2\` \u2014 **[\u6280\u80FD\u540D\u79F0]**\uFF1A[\u4E00\u53E5\u8BDD\u8BF4\u660E]
+>
+> \u662F\u5426\u6FC0\u6D3B\u8FD9\u4E9B\u6280\u80FD\u5F00\u59CB\u5206\u6790\uFF1F
+
+**\u7B2C\u4E09\u6B65 \u2014 \u6FC0\u6D3B\u6280\u80FD\uFF0C\u8BFB\u53D6 SKILL.md**
+\u7528\u6237\u786E\u8BA4\u540E\u7ACB\u5373\u6267\u884C\uFF1A\`cat ${WORKFLOWS_DIR}/<skill-id>/SKILL.md\`
+\u4E25\u683C\u6309\u7167 SKILL.md \u7684\u6BCF\u4E00\u6B65\u6267\u884C\uFF0C\u4E0D\u8DF3\u8FC7\u3001\u4E0D\u81EA\u884C\u53D1\u6325\u3002
+
+**\u7B2C\u56DB\u6B65 \u2014 \u53EA\u95EE\u771F\u6B63\u5FC5\u8981\u7684\u6570\u636E\u95EE\u9898**
+SKILL.md \u8BFB\u5B8C\u540E\uFF0C\u53EA\u8BE2\u95EE\u6267\u884C\u6240\u9700\u7684\u5173\u952E\u4FE1\u606F\uFF08\u5982\u672C\u5730\u6587\u4EF6\u8DEF\u5F84\u3001\u6837\u672C\u5206\u7EC4\u540D\uFF09\u3002
+
+\u274C **\u7EDD\u5BF9\u7981\u6B62**\uFF1A\u5728\u5B8C\u6210\u7B2C\u4E00\u6B65\uFF08\u8BC6\u522B\u6280\u80FD\uFF09\u4E4B\u524D\uFF0C\u8BE2\u95EE\u901A\u7528\u6570\u636E\u6536\u96C6\u95EE\u9898\u3002
+\u4F8B\u5982\u4E0B\u5217\u63D0\u95EE\u65B9\u5F0F\u662F**\u9519\u8BEF\u7684**\uFF0C\u5FC5\u987B\u907F\u514D\uFF1A
+- "\u60A8\u6709 GEO \u6570\u636E\u96C6\u5417\uFF1F"
+- "\u6570\u636E\u6765\u6E90\u662F\u4EC0\u4E48\uFF1F"
+- "\u60A8\u7684\u6570\u636E\u683C\u5F0F\u662F\u4EC0\u4E48\uFF1F"
+- "\u8BF7\u63D0\u4F9B\u4EE5\u4E0B\u4FE1\u606F\uFF1A1. \u6570\u636E\u6765\u6E90 2. \u6570\u636E\u5185\u5BB9 ..."
+
+\u5982\u679C\u5B9E\u5728\u627E\u4E0D\u5230\u4EFB\u4F55\u5339\u914D\u6280\u80FD\uFF0C\u624D\u5141\u8BB8\u63D0\u95EE\u5E76\u8BF4\u660E\u7406\u7531\u3002
 
 ---
 
-## Workflow Library (21 Workflows)
+## Skill Library (22 Bioinformatics Skills)
 
-All workflows are at: **${WORKFLOWS_DIR}**
+All skills are at: **${WORKFLOWS_DIR}**
 
-For any workflow, read its guide first:
+For any skill, read its guide first:
 \`\`\`bash
-cat ${WORKFLOWS_DIR}/<workflow-id>/SKILL.md
+cat ${WORKFLOWS_DIR}/<skill-id>/SKILL.md
 \`\`\`
 
 ### \u{1F9EC} Transcriptomics
@@ -14677,7 +14704,7 @@ cat ${WORKFLOWS_DIR}/<workflow-id>/SKILL.md
 
 ---
 
-## Molecular Biology Tools
+## Molecular Biology Skills
 
 Python tools are at: **${TOOLS_DIR}**
 
@@ -14710,16 +14737,16 @@ ${dbSection ?? "\uFF08\u6682\u672A\u6CE8\u518C\u4EFB\u4F55\u6570\u636E\u5E93\u30
 
 ---
 
-## OpenClaw Medical Skills (979\u4E2A\u4E13\u79D1\u6280\u80FD)
+## Extended Skills (979\u4E2A)
 
-\u6280\u80FD\u5E93\u4F4D\u4E8E: **${SKILLS_DIR}**
+\u66F4\u591A\u5185\u7F6E\u6280\u80FD\u4F4D\u4E8E: **${SKILLS_DIR}**
 
 \u6BCF\u4E2A\u6280\u80FD\u76EE\u5F55\u5305\u542B\u4E00\u4E2A \`SKILL.md\` \u6587\u4EF6\uFF0C\u8BFB\u53D6\u65B9\u5F0F:
 \`\`\`bash
 cat ${SKILLS_DIR}/<skill-id>/SKILL.md
 \`\`\`
 
-**\u6280\u80FD\u6DB5\u76D6\u7684\u4E3B\u8981\u9886\u57DF**\uFF08\u7528\u6237\u5DF2\u901A\u8FC7 /sk \u547D\u4EE4\u52A0\u8F7D\u65F6\u81EA\u52A8\u6CE8\u5165\u4E0A\u4E0B\u6587\uFF09:
+**\u6DB5\u76D6\u9886\u57DF**\uFF08\u7528\u6237\u901A\u8FC7 /sk \u547D\u4EE4\u52A0\u8F7D\u540E\u81EA\u52A8\u6CE8\u5165\u4E0A\u4E0B\u6587\uFF09:
 - \u6587\u732E\u68C0\u7D22: pubmed-search, arxiv-search, bgpt-paper-search
 - \u7ED3\u6784\u751F\u7269\u5B66: alphafold, alphafold-database, bindcraft, binder-design
 - \u5355\u7EC6\u80DE: anndata, cellagent-annotation, scvi-tools
@@ -14728,7 +14755,9 @@ cat ${SKILLS_DIR}/<skill-id>/SKILL.md
 - \u6297\u4F53\u8BBE\u8BA1: antibody-design-agent, armored-cart-design-agent
 - \u4E34\u5E8A: clinical-*, ehr-fhir-integration, autonomous-oncology-agent
 
-\u5F53\u7528\u6237\u63D0\u5230\u67D0\u4E2A\u4E13\u79D1\u4EFB\u52A1\u65F6\uFF0C\u5EFA\u8BAE\u4ED6\u4EEC\u4F7F\u7528 **/sk <\u5173\u952E\u8BCD>** \u641C\u7D22\u5E76\u52A0\u8F7D\u5BF9\u5E94\u6280\u80FD\u6307\u5357\u3002
+\u7528\u6237\u4E5F\u53EF\u4EE5\u901A\u8FC7 **/install** \u5B89\u88C5\u81EA\u5B9A\u4E49\u6280\u80FD\uFF0C\u5B89\u88C5\u540E\u4E0E\u5185\u7F6E\u6280\u80FD\u5B8C\u5168\u7B49\u4EF7\u3002
+
+\u5F53\u7528\u6237\u63D0\u5230\u67D0\u4E2A\u4EFB\u52A1\u65F6\uFF0C\u5EFA\u8BAE\u4F7F\u7528 **/sk <\u5173\u952E\u8BCD>** \u641C\u7D22\u5E76\u52A0\u8F7D\u5BF9\u5E94\u6280\u80FD\u3002
 
 ---
 
@@ -14819,9 +14848,9 @@ message(sprintf("\u603B\u57FA\u56E0\u6570: %d | \u663E\u8457 DEG: %d (\u4E0A\u8C
 ## Script Execution Rules
 
 \u{1F6A8} **\u5173\u952E\u89C4\u5219\uFF1A**
-1. **\u4F18\u5148\u4F7F\u7528\u5DE5\u4F5C\u6D41\u811A\u672C**\uFF0C\u4E0D\u8981\u4ECE\u96F6\u5199\u4EE3\u7801
+1. **\u4F18\u5148\u4F7F\u7528\u6280\u80FD\u5185\u7F6E\u811A\u672C**\uFF0C\u4E0D\u8981\u4ECE\u96F6\u5199\u4EE3\u7801
 2. **\u811A\u672C\u5931\u8D25\u5904\u7406\u987A\u5E8F**: \u4FEE\u590D\u5E76\u91CD\u8BD5 \u2192 \u4FEE\u6539\u811A\u672C \u2192 \u9002\u914D\u65B9\u6848 \u2192 \u6700\u540E\u624D\u4ECE\u5934\u5199
-3. **\u4F7F\u7528\u76F8\u5BF9\u8DEF\u5F84**\uFF1A\u5728\u5DE5\u4F5C\u6D41\u76EE\u5F55\u5185\u7528 \`source("scripts/xxx.R")\` \u800C\u975E\u7EDD\u5BF9\u8DEF\u5F84
+3. **\u4F7F\u7528\u76F8\u5BF9\u8DEF\u5F84**\uFF1A\u5728\u6280\u80FD\u76EE\u5F55\u5185\u7528 \`source("scripts/xxx.R")\` \u800C\u975E\u7EDD\u5BF9\u8DEF\u5F84
 4. **\u9A8C\u8BC1\u6D88\u606F**\uFF1A\u6BCF\u6B65\u5B8C\u6210\u5E94\u770B\u5230 "\u2713" \u786E\u8BA4\u6D88\u606F\uFF1B\u770B\u4E0D\u5230\u8BF4\u660E\u6CA1\u7528\u811A\u672C
 
 ---
@@ -14838,25 +14867,25 @@ message(sprintf("\u603B\u57FA\u56E0\u6570: %d | \u663E\u8457 DEG: %d (\u4E0A\u8C
 \u4F7F\u7528 HPC \u5DE5\u5177\u63D0\u4EA4 AlphaFold \u4EFB\u52A1
 
 **\u7528\u6237\u8BF4 "\u627E DEG \u7684\u901A\u8DEF" \u2192**
-\u4F7F\u7528 functional-enrichment-from-degs \u5DE5\u4F5C\u6D41
+\u4F7F\u7528 functional-enrichment-from-degs \u6280\u80FD
 
 **\u7528\u6237\u8BF4 "\u8BBE\u8BA1 CRISPR guide RNA" \u2192**
 \u4F7F\u7528 molecular_biology.py \u7684 design_crispr_knockout_guides()
 
 **\u7528\u6237\u8BF4 "\u54EA\u4E9B\u57FA\u56E0\u5728\u80BF\u7624\u91CC\u8868\u8FBE\u91CF\u9AD8" / "\u627E\u4E0A\u8C03\u57FA\u56E0" \u2192**
-\u2192 \u5DEE\u5F02\u8868\u8FBE\u5206\u6790\uFF08DESeq2\uFF09\uFF0C\u4F7F\u7528 bulk-rnaseq-counts-to-de-deseq2 \u5DE5\u4F5C\u6D41
+\u2192 \u5DEE\u5F02\u8868\u8FBE\u5206\u6790\uFF08DESeq2\uFF09\uFF0C\u4F7F\u7528 bulk-rnaseq-counts-to-de-deseq2 \u6280\u80FD
 
 **\u7528\u6237\u8BF4 "\u5148\u505A\u5DEE\u5F02\u8868\u8FBE\uFF0C\u518D\u505A\u5BCC\u96C6\u5206\u6790" \u2192**
 \u2192 \u8BC6\u522B\u4E3A\u591A\u4EFB\u52A1\uFF1A\u4F9D\u6B21\u6267\u884C bulk-rnaseq-counts-to-de-deseq2 + functional-enrichment-from-degs
 
 **\u7528\u6237\u8BF4 "\u8FD9\u4E9B\u57FA\u56E0\u53C2\u4E0E\u4EC0\u4E48\u901A\u8DEF" / "\u57FA\u56E0\u529F\u80FD\u662F\u4EC0\u4E48" \u2192**
-\u2192 \u529F\u80FD\u5BCC\u96C6\u5206\u6790\uFF0C\u4F7F\u7528 functional-enrichment-from-degs \u5DE5\u4F5C\u6D41
+\u2192 \u529F\u80FD\u5BCC\u96C6\u5206\u6790\uFF0C\u4F7F\u7528 functional-enrichment-from-degs \u6280\u80FD
 
 **\u7528\u6237\u8BF4 "\u5206\u6790\u5355\u7EC6\u80DE\u6570\u636E" / "10X\u6570\u636E\u5206\u6790" \u2192**
 \u5148\u95EE\uFF1APython \u8FD8\u662F R\uFF1F\u2192 Scanpy \u6216 Seurat
 
 **\u7528\u6237\u8BF4 "\u753B\u751F\u5B58\u66F2\u7EBF" / "\u5206\u6790\u60A3\u8005\u9884\u540E" / "OS/PFS \u5206\u6790" \u2192**
-\u2192 \u751F\u5B58\u5206\u6790\uFF0C\u4F7F\u7528 survival-analysis-clinical \u5DE5\u4F5C\u6D41
+\u2192 \u751F\u5B58\u5206\u6790\uFF0C\u4F7F\u7528 survival-analysis-clinical \u6280\u80FD
 
 **\u7528\u6237\u8BF4 "\u5E2E\u6211\u5206\u6790 GSE12345" / "\u4E0B\u8F7D GEO \u6570\u636E" \u2192**
 \u2192 \u7ACB\u5373\u8C03\u7528 fetch_geo("GSE12345") \u83B7\u53D6\u5143\u6570\u636E\u548C\u4E0B\u8F7D\u4EE3\u7801\uFF0C\u65E0\u9700\u8BA9\u7528\u6237\u624B\u52A8\u4E0B\u8F7D`;
@@ -15149,7 +15178,7 @@ var SKILL_ROUTES = [
     id: "bulk-rnaseq-counts-to-de-deseq2",
     name: "DESeq2 \u5DEE\u5F02\u8868\u8FBE\u5206\u6790",
     category: "\u8F6C\u5F55\u7EC4",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       // exact tool names
       "deseq2",
@@ -15184,7 +15213,7 @@ var SKILL_ROUTES = [
     id: "bulk-omics-clustering",
     name: "\u6837\u672C / \u7279\u5F81\u805A\u7C7B",
     category: "\u8F6C\u5F55\u7EC4",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "wgcna\u805A\u7C7B",
       "\u5C42\u6B21\u805A\u7C7B",
@@ -15209,7 +15238,7 @@ var SKILL_ROUTES = [
     id: "scrnaseq-scanpy-core-analysis",
     name: "scRNA-seq (Scanpy / Python)",
     category: "\u5355\u7EC6\u80DE",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "scanpy",
       "scrna-seq",
@@ -15234,7 +15263,7 @@ var SKILL_ROUTES = [
     id: "scrnaseq-seurat-core-analysis",
     name: "scRNA-seq (Seurat / R)",
     category: "\u5355\u7EC6\u80DE",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "seurat",
       "r\u8BED\u8A00\u5355\u7EC6\u80DE",
@@ -15252,7 +15281,7 @@ var SKILL_ROUTES = [
     id: "spatial-transcriptomics",
     name: "\u7A7A\u95F4\u8F6C\u5F55\u7EC4",
     category: "\u5355\u7EC6\u80DE",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "\u7A7A\u95F4\u8F6C\u5F55\u7EC4",
       "spatial transcriptomics",
@@ -15274,7 +15303,7 @@ var SKILL_ROUTES = [
     id: "coexpression-network",
     name: "WGCNA \u5171\u8868\u8FBE\u7F51\u7EDC",
     category: "\u8F6C\u5F55\u7EC4",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "wgcna",
       "\u5171\u8868\u8FBE\u7F51\u7EDC",
@@ -15293,7 +15322,7 @@ var SKILL_ROUTES = [
     id: "functional-enrichment-from-degs",
     name: "GO / KEGG / GSEA \u5BCC\u96C6\u5206\u6790",
     category: "\u8F6C\u5F55\u7EC4",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "\u5BCC\u96C6\u5206\u6790",
       "go\u5206\u6790",
@@ -15318,14 +15347,26 @@ var SKILL_ROUTES = [
       "\u901A\u8DEF\u5BCC\u96C6",
       "\u57FA\u56E0\u96C6\u5206\u6790",
       "ora\u5206\u6790",
-      "gsea\u5206\u6790"
+      "gsea\u5206\u6790",
+      // mutation → pathway → phenotype
+      "\u7A81\u53D8\u901A\u8DEF",
+      "\u7A81\u53D8\u5F71\u54CD\u901A\u8DEF",
+      "\u7A81\u53D8\u4E0B\u6E38\u901A\u8DEF",
+      "\u6851\u57FA\u56FE\u901A\u8DEF",
+      "kras\u901A\u8DEF",
+      "ras-raf-mek\u901A\u8DEF",
+      "mapk\u901A\u8DEF",
+      "pi3k\u901A\u8DEF",
+      "\u80BA\u764C\u901A\u8DEF",
+      "\u9A71\u52A8\u57FA\u56E0\u901A\u8DEF",
+      "\u4FE1\u53F7\u901A\u8DEF\u5BCC\u96C6"
     ]
   },
   {
     id: "grn-pyscenic",
     name: "pySCENIC \u57FA\u56E0\u8C03\u63A7\u7F51\u7EDC",
     category: "\u5355\u7EC6\u80DE",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "pyscenic",
       "scenic",
@@ -15346,7 +15387,7 @@ var SKILL_ROUTES = [
     id: "genetic-variant-annotation",
     name: "\u9057\u4F20\u53D8\u5F02\u6CE8\u91CA",
     category: "\u57FA\u56E0\u7EC4",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "\u53D8\u5F02\u6CE8\u91CA",
       "variant annotation",
@@ -15364,14 +15405,43 @@ var SKILL_ROUTES = [
       "vcf\u6587\u4EF6\u5206\u6790",
       "\u9057\u4F20\u53D8\u5F02\u89E3\u6790",
       "\u81F4\u75C5\u6027\u8BC4\u4F30",
-      "\u53D8\u5F02\u5F71\u54CD\u9884\u6D4B"
+      "\u53D8\u5F02\u5F71\u54CD\u9884\u6D4B",
+      // specific oncology mutations (KRAS/EGFR/ALK etc.)
+      "kras\u7A81\u53D8",
+      "kras g12c",
+      "kras g12d",
+      "kras g12v",
+      "kras mutation",
+      "egfr\u7A81\u53D8",
+      "egfr l858r",
+      "egfr\u5916\u663E\u5B5019\u7F3A\u5931",
+      "alk\u878D\u5408",
+      "braf v600e",
+      "tp53\u7A81\u53D8",
+      "pik3ca\u7A81\u53D8",
+      // cancer contexts
+      "nsclc\u7A81\u53D8",
+      "\u80BA\u764C\u7A81\u53D8",
+      "\u975E\u5C0F\u7EC6\u80DE\u80BA\u764C\u7A81\u53D8",
+      "\u80BA\u817A\u764C\u7A81\u53D8",
+      "\u80BF\u7624\u4F53\u7EC6\u80DE\u7A81\u53D8",
+      "\u9A71\u52A8\u57FA\u56E0\u7A81\u53D8",
+      "\u764C\u75C7\u9A71\u52A8\u7A81\u53D8",
+      // mutation landscape
+      "\u7A81\u53D8\u9891\u7387",
+      "\u8DE8\u764C\u79CD\u7A81\u53D8",
+      "\u7A81\u53D8\u70ED\u56FE",
+      "\u7A81\u53D8\u666F\u89C2\u56FE",
+      "\u7011\u5E03\u56FE\u7A81\u53D8",
+      "mutation landscape",
+      "oncoprint"
     ]
   },
   {
     id: "gwas-to-function-twas",
     name: "GWAS \u2192 TWAS \u529F\u80FD\u89E3\u6790",
     category: "\u57FA\u56E0\u7EC4",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "gwas\u5206\u6790",
       "twas",
@@ -15394,7 +15464,7 @@ var SKILL_ROUTES = [
     id: "mendelian-randomization-twosamplemr",
     name: "\u5B5F\u5FB7\u5C14\u968F\u673A\u5316 (MR)",
     category: "\u7EDF\u8BA1",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "\u5B5F\u5FB7\u5C14\u968F\u673A\u5316",
       "mendelian randomization",
@@ -15415,7 +15485,7 @@ var SKILL_ROUTES = [
     id: "polygenic-risk-score-prs-catalog",
     name: "PRS \u591A\u57FA\u56E0\u98CE\u9669\u8BC4\u5206",
     category: "\u57FA\u56E0\u7EC4",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "prs\u8BC4\u5206",
       "polygenic risk score",
@@ -15433,7 +15503,7 @@ var SKILL_ROUTES = [
     id: "pooled-crispr-screens",
     name: "CRISPR \u6587\u5E93\u7B5B\u9009 (MAGeCK/BAGEL2)",
     category: "\u57FA\u56E0\u7EC4",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "crispr\u6587\u5E93\u7B5B\u9009",
       "crispr screen",
@@ -15455,7 +15525,7 @@ var SKILL_ROUTES = [
     id: "chip-atlas-peak-enrichment",
     name: "ChIP-seq \u5CF0\u503C\u5BCC\u96C6 (ChIP-Atlas)",
     category: "\u8868\u89C2\u57FA\u56E0\u7EC4",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "chip-atlas",
       "chip-seq\u5CF0\u503C\u5BCC\u96C6",
@@ -15474,7 +15544,7 @@ var SKILL_ROUTES = [
     id: "chip-atlas-diff-analysis",
     name: "ChIP-seq \u5DEE\u5F02\u7ED3\u5408\u5206\u6790",
     category: "\u8868\u89C2\u57FA\u56E0\u7EC4",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "chip\u5DEE\u5F02\u5206\u6790",
       "differential binding",
@@ -15491,7 +15561,7 @@ var SKILL_ROUTES = [
     id: "chip-atlas-target-genes",
     name: "ChIP-seq \u9776\u57FA\u56E0\u9274\u5B9A",
     category: "\u8868\u89C2\u57FA\u56E0\u7EC4",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "chip\u9776\u57FA\u56E0",
       "chip target gene",
@@ -15511,7 +15581,7 @@ var SKILL_ROUTES = [
     id: "clinicaltrials-landscape",
     name: "\u4E34\u5E8A\u8BD5\u9A8C\u683C\u5C40\u5206\u6790",
     category: "\u4E34\u5E8A",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "clinicaltrials\u5206\u6790",
       "clinical trial landscape",
@@ -15529,7 +15599,7 @@ var SKILL_ROUTES = [
     id: "literature-preclinical",
     name: "\u4E34\u5E8A\u524D\u6587\u732E\u7CFB\u7EDF\u63D0\u53D6",
     category: "\u6587\u732E",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "\u4E34\u5E8A\u524D\u6587\u732E",
       "preclinical literature",
@@ -15547,7 +15617,7 @@ var SKILL_ROUTES = [
     id: "experimental-design-statistics",
     name: "\u5B9E\u9A8C\u8BBE\u8BA1\u4E0E\u7EDF\u8BA1\u68C0\u9A8C",
     category: "\u7EDF\u8BA1",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "\u6837\u672C\u91CF\u8BA1\u7B97",
       "\u7EDF\u8BA1\u68C0\u9A8C\u9009\u62E9",
@@ -15573,7 +15643,7 @@ var SKILL_ROUTES = [
     id: "lasso-biomarker-panel",
     name: "LASSO \u751F\u7269\u6807\u5FD7\u7269\u9762\u677F\u7B5B\u9009",
     category: "\u7EDF\u8BA1",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "lasso\u56DE\u5F52\u7B5B\u9009",
       "lasso\u751F\u7269\u6807\u5FD7\u7269",
@@ -15594,7 +15664,7 @@ var SKILL_ROUTES = [
     id: "pcr-primer-design",
     name: "PCR / qPCR \u5F15\u7269\u8BBE\u8BA1",
     category: "\u5206\u5B50\u751F\u7269\u5B66",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       "\u5F15\u7269\u8BBE\u8BA1",
       "primer design",
@@ -15616,7 +15686,7 @@ var SKILL_ROUTES = [
     id: "survival-analysis-clinical",
     name: "\u4E34\u5E8A\u751F\u5B58\u5206\u6790 (KM + Cox)",
     category: "\u4E34\u5E8A",
-    tag: "workflow",
+    tag: "builtin",
     keywords: [
       // exact method names
       "kaplan-meier",
@@ -15663,12 +15733,12 @@ var SKILL_ROUTES = [
       "\u7A81\u53D8\u4E0E\u9884\u540E"
     ]
   },
-  // ── OpenClaw Key Skills ────────────────────────────────────────────────────────
+  // ── Extended Skills (key routable entries) ───────────────────────────────────
   {
     id: "pubmed-search",
     name: "PubMed \u6587\u732E\u68C0\u7D22",
     category: "\u6587\u732E",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "pubmed\u68C0\u7D22",
       "pubmed\u641C\u7D22",
@@ -15683,14 +15753,14 @@ var SKILL_ROUTES = [
     id: "arxiv-search",
     name: "arXiv \u9884\u5370\u672C\u68C0\u7D22",
     category: "\u6587\u732E",
-    tag: "skill",
+    tag: "builtin",
     keywords: ["arxiv\u68C0\u7D22", "arxiv\u641C\u7D22", "\u9884\u5370\u672C\u68C0\u7D22", "\u9884\u5370\u672C\u8BBA\u6587", "arxiv\u6587\u732E"]
   },
   {
     id: "alphafold",
     name: "AlphaFold \u86CB\u767D\u8D28\u7ED3\u6784\u9884\u6D4B",
     category: "\u7ED3\u6784\u751F\u7269\u5B66",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "alphafold\u7ED3\u6784\u9884\u6D4B",
       "alphafold\u8FD0\u884C",
@@ -15704,7 +15774,7 @@ var SKILL_ROUTES = [
     id: "alphafold-database",
     name: "AlphaFold \u6570\u636E\u5E93\u67E5\u8BE2",
     category: "\u7ED3\u6784\u751F\u7269\u5B66",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "alphafold\u6570\u636E\u5E93",
       "alphafold db",
@@ -15717,7 +15787,7 @@ var SKILL_ROUTES = [
     id: "bindcraft",
     name: "BindCraft \u86CB\u767D\u8D28\u7ED3\u5408\u4F53\u8BBE\u8BA1",
     category: "\u7ED3\u6784\u751F\u7269\u5B66",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "bindcraft",
       "\u7ED3\u5408\u4F53\u8BBE\u8BA1",
@@ -15731,7 +15801,7 @@ var SKILL_ROUTES = [
     id: "anndata",
     name: "AnnData \u5355\u7EC6\u80DE\u6570\u636E\u64CD\u4F5C",
     category: "\u5355\u7EC6\u80DE",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "anndata\u64CD\u4F5C",
       "h5ad\u6587\u4EF6\u5904\u7406",
@@ -15745,7 +15815,7 @@ var SKILL_ROUTES = [
     id: "cellagent-annotation",
     name: "CellAgent \u7EC6\u80DE\u7C7B\u578B\u81EA\u52A8\u6CE8\u91CA",
     category: "\u5355\u7EC6\u80DE",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "\u7EC6\u80DE\u7C7B\u578B\u6CE8\u91CA",
       "cell type annotation",
@@ -15759,7 +15829,7 @@ var SKILL_ROUTES = [
     id: "scvi-tools",
     name: "scVI \u5355\u7EC6\u80DE\u6DF1\u5EA6\u5B66\u4E60",
     category: "\u5355\u7EC6\u80DE",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "scvi-tools",
       "scvi\u6A21\u578B",
@@ -15773,21 +15843,54 @@ var SKILL_ROUTES = [
     id: "agentd-drug-discovery",
     name: "\u836F\u7269\u53D1\u73B0 Agent",
     category: "\u836F\u7269",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "\u836F\u7269\u53D1\u73B0\u6D41\u7A0B",
       "drug discovery",
       "\u5019\u9009\u836F\u7269\u7B5B\u9009",
       "\u5148\u5BFC\u5316\u5408\u7269",
       "hit\u5316\u5408\u7269",
-      "\u836F\u7269\u9776\u70B9\u53D1\u73B0"
+      "\u836F\u7269\u9776\u70B9\u53D1\u73B0",
+      // drug target analysis
+      "\u9776\u70B9\u8BC4\u4F30",
+      "\u9776\u70B9\u53EF\u89C6\u5316",
+      "\u9776\u70B9\u62A5\u544A",
+      "\u9776\u70B9\u5206\u6790",
+      "\u9776\u70B9\u53EF\u6210\u836F\u6027",
+      "druggability",
+      "\u53EF\u6210\u836F\u6027\u8BC4\u4F30",
+      "\u9776\u70B9\u7EFC\u5408\u8BC4\u5206",
+      "target visualization",
+      "drug target report",
+      "target assessment",
+      // specific target contexts
+      "kras\u9776\u70B9",
+      "kras\u53EF\u6210\u836F\u6027",
+      "kras\u6291\u5236\u5242",
+      "kras\u7A81\u53D8\u9776\u70B9",
+      "egfr\u9776\u70B9",
+      "alk\u9776\u70B9",
+      "braf\u9776\u70B9",
+      "her2\u9776\u70B9",
+      "met\u9776\u70B9",
+      // visualization types for target reports
+      "\u96F7\u8FBE\u56FE\u8BC4\u5206",
+      "\u6C14\u6CE1\u56FE\u9776\u70B9",
+      "\u6851\u57FA\u56FE\u7A81\u53D8\u901A\u8DEF",
+      "\u9776\u70B9\u70ED\u56FE",
+      // natural-language synonyms
+      "\u54EA\u4E9B\u9776\u70B9\u503C\u5F97\u5F00\u53D1",
+      "\u9776\u70B9\u4F18\u5148\u7EA7",
+      "\u9776\u70B9\u6210\u836F\u6027\u5206\u6790",
+      "\u9776\u5411\u6CBB\u7597\u9776\u70B9",
+      "\u7CBE\u51C6\u533B\u5B66\u9776\u70B9"
     ]
   },
   {
     id: "chembl-database",
     name: "ChEMBL \u5316\u5408\u7269\u6D3B\u6027\u6570\u636E\u5E93",
     category: "\u836F\u7269",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "chembl\u67E5\u8BE2",
       "chembl\u6570\u636E\u5E93",
@@ -15801,7 +15904,7 @@ var SKILL_ROUTES = [
     id: "rdkit",
     name: "RDKit \u5316\u5B66\u4FE1\u606F\u5B66",
     category: "\u836F\u7269",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "rdkit",
       "\u5316\u5B66\u4FE1\u606F\u5B66",
@@ -15816,7 +15919,7 @@ var SKILL_ROUTES = [
     id: "bio-variant-calling",
     name: "\u53D8\u5F02\u68C0\u6D4B\u6D41\u7A0B (GATK/Mutect2)",
     category: "\u57FA\u56E0\u7EC4",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "\u53D8\u5F02\u68C0\u6D4B\u6D41\u7A0B",
       "variant calling\u6D41\u7A0B",
@@ -15831,7 +15934,7 @@ var SKILL_ROUTES = [
     id: "antibody-design-agent",
     name: "\u6297\u4F53\u8BBE\u8BA1 Agent",
     category: "\u6297\u4F53",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "\u6297\u4F53\u8BBE\u8BA1",
       "antibody design",
@@ -15845,7 +15948,7 @@ var SKILL_ROUTES = [
     id: "armored-cart-design-agent",
     name: "Armored CAR-T \u8BBE\u8BA1",
     category: "\u6297\u4F53",
-    tag: "skill",
+    tag: "builtin",
     keywords: [
       "car-t\u8BBE\u8BA1",
       "cart\u7EC6\u80DE\u8BBE\u8BA1",
@@ -16001,7 +16104,7 @@ function clearCheckpoints(sessionId) {
 
 // src/index.ts
 var import_fs7 = require("fs");
-var VERSION2 = "2.4.3";
+var VERSION2 = "2.4.4";
 var SKILLHUB_HUBS = {
   bgi: { label: "BGI\u5185\u7F51", apiBase: "https://clawhub.ai", backend: "clawhub" },
   clawhub: { label: "clawhub.ai", apiBase: "https://clawhub.ai", backend: "clawhub" },
@@ -16151,8 +16254,8 @@ function installBundledData() {
   const installedDataVersion = (0, import_fs6.existsSync)(DATA_VERSION_FILE) ? (0, import_fs6.readFileSync)(DATA_VERSION_FILE, "utf8").trim() : "";
   const needsUpdate = installedDataVersion !== VERSION2;
   const targets = [
-    { src: (0, import_path6.join)(bundledData, "workflows"), dest: WORKFLOWS_DIR, name: "Skills (\u751F\u4FE1\u5DE5\u4F5C\u6D41)" },
-    { src: (0, import_path6.join)(bundledData, "skills"), dest: SKILLS_DIR, name: "Skills (\u533B\u5B66\u4E13\u79D1)" },
+    { src: (0, import_path6.join)(bundledData, "workflows"), dest: WORKFLOWS_DIR, name: "Skills (\u5206\u6790\u7C7B)" },
+    { src: (0, import_path6.join)(bundledData, "skills"), dest: SKILLS_DIR, name: "Skills (\u6269\u5C55\u7C7B)" },
     { src: (0, import_path6.join)(bundledData, "tools"), dest: TOOLS_DIR, name: "\u5DE5\u5177" }
   ];
   let installed = false;
@@ -16231,7 +16334,7 @@ function printHelp() {
   console.log(`  ${source_default.cyan("!<\u547D\u4EE4>")}             \u7ED5\u8FC7 AI \u76F4\u63A5\u6267\u884C Shell \u547D\u4EE4\uFF08\u5B9E\u65F6\u8F93\u51FA\uFF09`);
   console.log(source_default.dim("  \u793A\u4F8B: !ls -la  !Rscript analysis.R  !python script.py  !samtools view -h a.bam"));
   console.log();
-  console.log(source_default.bold.cyan("\u2500\u2500\u2500 \u5DE5\u4F5C\u6D41\u5411\u5BFC \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500"));
+  console.log(source_default.bold.cyan("\u2500\u2500\u2500 Skill \u5411\u5BFC \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500"));
   console.log(`  ${source_default.cyan("/run")} <skill-id>     \u4EA4\u4E92\u5F0F\u53C2\u6570\u5411\u5BFC\uFF0C\u81EA\u52A8\u751F\u6210\u5E76\u6267\u884C\u5206\u6790\u811A\u672C`);
   console.log(`  ${source_default.cyan("/check-env")} [id]     \u68C0\u6D4B Skill \u6240\u9700 R/Python \u5305\u662F\u5426\u5DF2\u5B89\u88C5`);
   console.log(`  ${source_default.cyan("/search")} <\u5173\u952E\u8BCD>    \u5728 SkillHub \u641C\u7D22\u5E76\u4E0B\u8F7D\u6280\u80FD ${source_default.dim("[--hub=bgi|clawhub|tencent]")}`);
@@ -16976,7 +17079,7 @@ async function handleCommand(input, rl, history, thinkMode, injectedSkills) {
       }
       break;
     }
-    // ── /run workflow wizard ─────────────────────────────────────────────────
+    // ── /run skill wizard ────────────────────────────────────────────────────
     case "run": {
       const targetId = arg;
       if (!targetId) {
@@ -17017,7 +17120,7 @@ async function handleCommand(input, rl, history, thinkMode, injectedSkills) {
         if (ans.trim()) answers[q2] = ans.trim();
       }
       const paramSummary = Object.entries(answers).map(([k2, v2]) => `- ${k2}: ${v2}`).join("\n");
-      const runPrompt = `\u8BF7\u6839\u636E\u4EE5\u4E0B\u53C2\u6570\uFF0C\u4F7F\u7528 ${skillName || runMatch.id} \u5DE5\u4F5C\u6D41\u751F\u6210\u5B8C\u6574\u7684\u5206\u6790\u811A\u672C\u5E76\u7ACB\u5373\u6267\u884C\uFF1A
+      const runPrompt = `\u8BF7\u6839\u636E\u4EE5\u4E0B\u53C2\u6570\uFF0C\u4F7F\u7528 ${skillName || runMatch.id} \u6280\u80FD\u751F\u6210\u5B8C\u6574\u7684\u5206\u6790\u811A\u672C\u5E76\u7ACB\u5373\u6267\u884C\uFF1A
 
 \u7528\u6237\u63D0\u4F9B\u7684\u53C2\u6570\uFF1A
 ${paramSummary}
@@ -17878,16 +17981,29 @@ async function main() {
     const { routes: suggestedRoutes, topScore } = routeSkill(trimmed);
     const newRoutes = suggestedRoutes.filter((r2) => !injectedSkills.has(r2.id));
     if (newRoutes.length === 1 && topScore >= 8) {
-      const r2 = newRoutes[0];
-      const ok = await injectSkill(r2.id, history, injectedSkills, rl, true);
-      if (ok) {
-        console.log(source_default.dim("  (\u63D0\u793A: /clear \u53EF\u6E05\u9664\u4E0A\u4E0B\u6587\u540E\u5207\u6362 Skill)"));
-      }
-    } else if (newRoutes.length >= 2 && topScore >= 4) {
-      console.log(source_default.dim("\n\u{1F4A1} \u68C0\u6D4B\u5230\u76F8\u5173\u6280\u80FD\uFF0C\u8F93\u5165 /sk <id> \u6FC0\u6D3B:"));
-      newRoutes.slice(0, 3).forEach((r2) => {
-        console.log(source_default.dim(`   /sk ${r2.id}  \u2014 ${r2.name}`));
+      await injectSkill(newRoutes[0].id, history, injectedSkills, rl, false);
+    } else if (newRoutes.length >= 1 && topScore >= 2) {
+      const candidates = newRoutes.slice(0, 3);
+      console.log();
+      console.log(source_default.bold.yellow("\u250C\u2500 \u{1F50D} \u68C0\u6D4B\u5230\u76F8\u5173\u6280\u80FD \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500"));
+      candidates.forEach((r2, i2) => {
+        console.log(`\u2502  ${source_default.bold.cyan(`[${i2 + 1}]`)} ${source_default.cyan(r2.id)}${source_default.dim("  \u2014")} ${r2.name}`);
       });
+      console.log(source_default.bold.yellow("\u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500"));
+      const pickPrompt = candidates.length > 1 ? `  \u6FC0\u6D3B\u54EA\u4E2A\u6280\u80FD\uFF1F\u8F93\u5165\u7F16\u53F7 (\u5982 1 \u6216 1,2)\uFF0C\u56DE\u8F66\u8DF3\u8FC7 \u203A ` : `  \u6FC0\u6D3B\u6B64\u6280\u80FD\uFF1F[Y/n] \u203A `;
+      const skillChoice = await question(rl, source_default.yellow(pickPrompt));
+      const choiceTrimmed = skillChoice.trim().toLowerCase();
+      if (choiceTrimmed !== "" && choiceTrimmed !== "n") {
+        let indices;
+        if (candidates.length === 1) {
+          indices = choiceTrimmed === "y" || choiceTrimmed === "" ? [0] : [];
+        } else {
+          indices = choiceTrimmed.split(",").map((s2) => parseInt(s2.trim(), 10) - 1).filter((i2) => i2 >= 0 && i2 < candidates.length);
+        }
+        for (const idx of indices) {
+          await injectSkill(candidates[idx].id, history, injectedSkills, rl, true);
+        }
+      }
       console.log();
     }
     const expanded = expandFileRefs(trimmed);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bgicli/bgicli",
-  "version": "2.4.3",
+  "version": "2.4.5",
   "description": "BGI CLI — Bioinformatics AI terminal for Chinese researchers (百炼/DeepSeek/Kimi/Qwen)",
   "bin": {
     "bgi": "dist/bgi.js"