chainlesschain 0.43.2 → 0.44.0

package/src/commands/init.js

@@ -11,6 +11,2106 @@ import path from "path";
 import { logger } from "../lib/logger.js";
 import { isInsideProject, findProjectRoot } from "../lib/project-detector.js";
 
+// Workspace skill templates for ai-media-creator
+const SKILL_TEMPLATES = {
+  "comfyui-image": {
+    md: `---
+name: comfyui-image
+display-name: ComfyUI 图像生成
+category: media
+description: 通过 ComfyUI REST API 生成图像（文生图/图生图），支持自定义工作流
+version: 1.0.0
+author: ChainlessChain
+parameters:
+  - name: prompt
+    type: string
+    required: true
+    description: 图像生成提示词（正向）
+  - name: negative_prompt
+    type: string
+    required: false
+    description: 负向提示词
+    default: ""
+  - name: width
+    type: number
+    required: false
+    description: 图像宽度（像素）
+    default: 512
+  - name: height
+    type: number
+    required: false
+    description: 图像高度（像素）
+    default: 512
+  - name: steps
+    type: number
+    required: false
+    description: 采样步数
+    default: 20
+  - name: workflow
+    type: string
+    required: false
+    description: 自定义工作流 JSON 文件路径（位于 workflows/ 目录）
+execution-mode: direct
+---
+
+# ComfyUI 图像生成
+
+通过本地 ComfyUI 服务（默认 http://localhost:8188）生成 AI 图像。
+
+## 使用示例
+
+\`\`\`bash
+# 简单文生图
+chainlesschain skill run comfyui-image "a sunset over mountains, oil painting style"
+
+# 指定尺寸和步数
+chainlesschain skill run comfyui-image "portrait of a warrior" --args '{"width":768,"height":1024,"steps":30}'
+
+# 使用自定义工作流
+chainlesschain skill run comfyui-image "cyberpunk city" --args '{"workflow":"workflows/my-workflow.json"}'
+\`\`\`
+
+## 前提条件
+
+- ComfyUI 已安装并运行（默认端口 8188）
+- 至少加载了一个 Stable Diffusion 模型
+- 安装了 ComfyUI 的 SaveImage 节点（默认已包含）
+
+## cli-anything 集成说明
+
+如果你有带 CLI 接口的 AI 图像工具（如第三方 ComfyUI CLI 包装、InvokeAI CLI 等），可以通过以下方式注册：
+
+\`\`\`bash
+chainlesschain cli-anything register <tool-name>
+\`\`\`
+`,
+    handler: `/**
+ * ComfyUI Image Generation Skill Handler
+ * Calls ComfyUI REST API to generate images via Stable Diffusion workflows.
+ *
+ * Requirements: ComfyUI running at COMFYUI_URL (default: http://localhost:8188)
+ */
+
+const http = require("http");
+const https = require("https");
+const fs = require("fs");
+const path = require("path");
+const url = require("url");
+
+const COMFYUI_URL = process.env.COMFYUI_URL || "http://localhost:8188";
+const POLL_INTERVAL_MS = 2000;
+const POLL_TIMEOUT_MS = 120000;
+
+function httpRequest(urlStr, options = {}, body = null) {
+  return new Promise((resolve, reject) => {
+    const parsed = url.parse(urlStr);
+    const lib = parsed.protocol === "https:" ? https : http;
+    const reqOpts = {
+      hostname: parsed.hostname,
+      port: parsed.port,
+      path: parsed.path,
+      method: options.method || "GET",
+      headers: options.headers || {},
+    };
+    const req = lib.request(reqOpts, (res) => {
+      let data = "";
+      res.on("data", (chunk) => (data += chunk.toString("utf8")));
+      res.on("end", () => {
+        try {
+          resolve({ status: res.statusCode, body: JSON.parse(data) });
+        } catch {
+          resolve({ status: res.statusCode, body: data });
+        }
+      });
+    });
+    req.on("error", reject);
+    if (body) req.write(typeof body === "string" ? body : JSON.stringify(body));
+    req.end();
+  });
+}
+
+function buildDefaultWorkflow(params) {
+  const { prompt, negative_prompt, width, height, steps } = params;
+  return {
+    "1": {
+      class_type: "CheckpointLoaderSimple",
+      inputs: { ckpt_name: "v1-5-pruned-emaonly.ckpt" },
+    },
+    "2": {
+      class_type: "EmptyLatentImage",
+      inputs: { width: width || 512, height: height || 512, batch_size: 1 },
+    },
+    "3": {
+      class_type: "CLIPTextEncode",
+      inputs: { text: prompt, clip: ["1", 1] },
+    },
+    "4": {
+      class_type: "CLIPTextEncode",
+      inputs: { text: negative_prompt || "", clip: ["1", 1] },
+    },
+    "5": {
+      class_type: "KSampler",
+      inputs: {
+        model: ["1", 0],
+        positive: ["3", 0],
+        negative: ["4", 0],
+        latent_image: ["2", 0],
+        seed: Math.floor(Math.random() * 2 ** 32),
+        steps: steps || 20,
+        cfg: 7,
+        sampler_name: "euler",
+        scheduler: "normal",
+        denoise: 1,
+      },
+    },
+    "6": {
+      class_type: "VAEDecode",
+      inputs: { samples: ["5", 0], vae: ["1", 2] },
+    },
+    "7": {
+      class_type: "SaveImage",
+      inputs: { images: ["6", 0], filename_prefix: "cc_" },
+    },
+  };
+}
+
+async function pollUntilDone(promptId) {
+  const start = Date.now();
+  while (Date.now() - start < POLL_TIMEOUT_MS) {
+    await new Promise((r) => setTimeout(r, POLL_INTERVAL_MS));
+    const res = await httpRequest(\`\${COMFYUI_URL}/history/\${promptId}\`);
+    if (res.status === 200 && res.body && res.body[promptId]) {
+      const hist = res.body[promptId];
+      if (hist.status && hist.status.completed) {
+        // Collect output images
+        const images = [];
+        for (const nodeId of Object.keys(hist.outputs || {})) {
+          const nodeOut = hist.outputs[nodeId];
+          if (nodeOut.images) {
+            for (const img of nodeOut.images) {
+              images.push({
+                filename: img.filename,
+                subfolder: img.subfolder || "",
+                url: \`\${COMFYUI_URL}/view?filename=\${encodeURIComponent(img.filename)}&subfolder=\${encodeURIComponent(img.subfolder || "")}&type=output\`,
+              });
+            }
+          }
+        }
+        return { success: true, images };
+      }
+      if (hist.status && hist.status.status_str === "error") {
+        return { success: false, error: "ComfyUI reported workflow error" };
+      }
+    }
+  }
+  return { success: false, error: "Timeout waiting for image generation" };
+}
+
+async function comfyuiImageHandler(params) {
+  const { prompt, negative_prompt, width, height, steps, workflow } = params;
+
+  if (!prompt) {
+    return { error: "Missing required parameter: prompt" };
+  }
+
+  // Check ComfyUI is running
+  let statsRes;
+  try {
+    statsRes = await httpRequest(\`\${COMFYUI_URL}/system_stats\`);
+  } catch (err) {
+    return {
+      error: \`Cannot connect to ComfyUI at \${COMFYUI_URL}. Is it running?\`,
+      hint: "Start ComfyUI first: python main.py --listen 0.0.0.0",
+    };
+  }
+  if (statsRes.status !== 200) {
+    return { error: \`ComfyUI returned HTTP \${statsRes.status}\` };
+  }
+
+  // Load workflow
+  let workflowJson;
+  if (workflow) {
+    const wfPath = path.isAbsolute(workflow)
+      ? workflow
+      : path.join(process.cwd(), workflow);
+    if (!fs.existsSync(wfPath)) {
+      return { error: \`Workflow file not found: \${wfPath}\` };
+    }
+    try {
+      workflowJson = JSON.parse(fs.readFileSync(wfPath, "utf-8"));
+    } catch (err) {
+      return { error: \`Failed to parse workflow JSON: \${err.message}\` };
+    }
+  } else {
+    workflowJson = buildDefaultWorkflow({ prompt, negative_prompt, width, height, steps });
+  }
+
+  // Submit prompt
+  const submitRes = await httpRequest(
+    \`\${COMFYUI_URL}/prompt\`,
+    {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+    },
+    { prompt: workflowJson },
+  );
+
+  if (submitRes.status !== 200 || !submitRes.body || !submitRes.body.prompt_id) {
+    return {
+      error: \`Failed to submit workflow: HTTP \${submitRes.status}\`,
+      detail: submitRes.body,
+    };
+  }
+
+  const promptId = submitRes.body.prompt_id;
+  console.log(\`[comfyui-image] Submitted prompt \${promptId}, waiting...\`);
+
+  // Poll for result
+  const result = await pollUntilDone(promptId);
+  if (!result.success) {
+    return { error: result.error };
+  }
+
+  return {
+    success: true,
+    promptId,
+    images: result.images,
+    message: \`Generated \${result.images.length} image(s). Open URLs to download:\`,
+    urls: result.images.map((i) => i.url),
+  };
+}
+
+comfyuiImageHandler.execute = async (task, _ctx, _skill) => {
+  const input = typeof task === "string" ? task : (task.input || task.params?.input || "");
+  let p = {};
+  try { p = input.trim().startsWith("{") ? JSON.parse(input) : { prompt: input }; }
+  catch { p = { prompt: input }; }
+  return comfyuiImageHandler(p);
+};
+module.exports = comfyuiImageHandler;
+`,
+  },
+  "comfyui-video": {
+    md: `---
+name: comfyui-video
+display-name: ComfyUI 视频/动画生成
+category: media
+description: 通过 ComfyUI + AnimateDiff 节点生成 AI 动画视频，支持自定义工作流
+version: 1.0.0
+author: ChainlessChain
+parameters:
+  - name: prompt
+    type: string
+    required: true
+    description: 视频内容描述提示词
+  - name: frames
+    type: number
+    required: false
+    description: 帧数（AnimateDiff 推荐 16 的倍数）
+    default: 16
+  - name: fps
+    type: number
+    required: false
+    description: 输出帧率
+    default: 8
+  - name: workflow
+    type: string
+    required: true
+    description: AnimateDiff 工作流 JSON 文件路径（位于 workflows/ 目录）
+execution-mode: direct
+---
+
+# ComfyUI 视频/动画生成
+
+使用 ComfyUI + AnimateDiff 扩展生成 AI 动画视频。
+
+## 前提条件
+
+1. ComfyUI 已安装并运行（默认端口 8188）
+2. 安装 AnimateDiff ComfyUI 扩展：
+   \`\`\`bash
+   cd ComfyUI/custom_nodes
+   git clone https://github.com/guoyww/AnimateDiff-EvolutionaryFramework
+   \`\`\`
+3. 下载 AnimateDiff 模型至 \`ComfyUI/models/animatediff_models/\`
+4. 准备工作流 JSON 文件（保存至 \`workflows/\` 目录）
+
+## 使用示例
+
+\`\`\`bash
+# 使用自定义 AnimateDiff 工作流
+chainlesschain skill run comfyui-video "a cat walking" --args '{"workflow":"workflows/animatediff.json","frames":16}'
+\`\`\`
+
+## 工作流说明
+
+由于 AnimateDiff 工作流需要针对具体模型定制，本技能要求提供工作流文件。
+在 workflows/README.md 中可以找到工作流模板说明。
+
+## cli-anything 集成
+
+如果有支持 CLI 的视频生成工具（如 deforum-cli、svd-cli 等），可以通过以下命令注册：
+
+\`\`\`bash
+chainlesschain cli-anything register <tool-name>
+\`\`\`
+`,
+    handler: `/**
+ * ComfyUI Video/Animation Generation Skill Handler
+ * Uses ComfyUI + AnimateDiff extension to generate animated videos.
+ *
+ * Requirements:
+ *   - ComfyUI running at COMFYUI_URL (default: http://localhost:8188)
+ *   - AnimateDiff custom node installed
+ *   - A workflow JSON file (required — provide via 'workflow' parameter)
+ */
+
+const http = require("http");
+const https = require("https");
+const fs = require("fs");
+const path = require("path");
+const url = require("url");
+
+const COMFYUI_URL = process.env.COMFYUI_URL || "http://localhost:8188";
+const POLL_INTERVAL_MS = 3000;
+const POLL_TIMEOUT_MS = 300000; // 5 min for video
+
+function httpRequest(urlStr, options = {}, body = null) {
+  return new Promise((resolve, reject) => {
+    const parsed = url.parse(urlStr);
+    const lib = parsed.protocol === "https:" ? https : http;
+    const reqOpts = {
+      hostname: parsed.hostname,
+      port: parsed.port,
+      path: parsed.path,
+      method: options.method || "GET",
+      headers: options.headers || {},
+    };
+    const req = lib.request(reqOpts, (res) => {
+      let data = "";
+      res.on("data", (chunk) => (data += chunk.toString("utf8")));
+      res.on("end", () => {
+        try {
+          resolve({ status: res.statusCode, body: JSON.parse(data) });
+        } catch {
+          resolve({ status: res.statusCode, body: data });
+        }
+      });
+    });
+    req.on("error", reject);
+    if (body) req.write(typeof body === "string" ? body : JSON.stringify(body));
+    req.end();
+  });
+}
+
+async function pollUntilDone(promptId) {
+  const start = Date.now();
+  while (Date.now() - start < POLL_TIMEOUT_MS) {
+    await new Promise((r) => setTimeout(r, POLL_INTERVAL_MS));
+    const res = await httpRequest(\`\${COMFYUI_URL}/history/\${promptId}\`);
+    if (res.status === 200 && res.body && res.body[promptId]) {
+      const hist = res.body[promptId];
+      if (hist.status && hist.status.completed) {
+        const outputs = [];
+        for (const nodeId of Object.keys(hist.outputs || {})) {
+          const nodeOut = hist.outputs[nodeId];
+          // Collect video/gif/image outputs
+          for (const key of ["videos", "gifs", "images"]) {
+            if (nodeOut[key]) {
+              for (const item of nodeOut[key]) {
+                outputs.push({
+                  type: key,
+                  filename: item.filename,
+                  subfolder: item.subfolder || "",
+                  url: \`\${COMFYUI_URL}/view?filename=\${encodeURIComponent(item.filename)}&subfolder=\${encodeURIComponent(item.subfolder || "")}&type=output\`,
+                });
+              }
+            }
+          }
+        }
+        return { success: true, outputs };
+      }
+      if (hist.status && hist.status.status_str === "error") {
+        return { success: false, error: "ComfyUI reported workflow error" };
+      }
+    }
+  }
+  return { success: false, error: "Timeout waiting for video generation" };
+}
+
+async function comfyuiVideoHandler(params) {
+  const { prompt, frames, fps, workflow } = params;
+
+  if (!prompt) {
+    return { error: "Missing required parameter: prompt" };
+  }
+
+  if (!workflow) {
+    return {
+      error: "Missing required parameter: workflow",
+      hint: [
+        "AnimateDiff requires a custom workflow JSON file.",
+        "1. Create your AnimateDiff workflow in ComfyUI UI",
+        "2. Save it to workflows/ directory in this project",
+        "3. Run: chainlesschain skill run comfyui-video \\"your prompt\\" --args \\'{}\\"workflow\\":\\"workflows/your-workflow.json\\"}\\' ",
+        "",
+        "See workflows/README.md for workflow template guidance.",
+      ].join("\\n"),
+    };
+  }
+
+  // Check ComfyUI is running
+  try {
+    const statsRes = await httpRequest(\`\${COMFYUI_URL}/system_stats\`);
+    if (statsRes.status !== 200) {
+      throw new Error(\`HTTP \${statsRes.status}\`);
+    }
+  } catch (err) {
+    return {
+      error: \`Cannot connect to ComfyUI at \${COMFYUI_URL}. Is it running?\`,
+      hint: "Start ComfyUI first: python main.py --listen 0.0.0.0",
+    };
+  }
+
+  // Load workflow file
+  const wfPath = path.isAbsolute(workflow)
+    ? workflow
+    : path.join(process.cwd(), workflow);
+  if (!fs.existsSync(wfPath)) {
+    return { error: \`Workflow file not found: \${wfPath}\` };
+  }
+
+  let workflowJson;
+  try {
+    workflowJson = JSON.parse(fs.readFileSync(wfPath, "utf-8"));
+  } catch (err) {
+    return { error: \`Failed to parse workflow JSON: \${err.message}\` };
+  }
+
+  // Submit prompt
+  const submitRes = await httpRequest(
+    \`\${COMFYUI_URL}/prompt\`,
+    {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+    },
+    { prompt: workflowJson },
+  );
+
+  if (submitRes.status !== 200 || !submitRes.body || !submitRes.body.prompt_id) {
+    return {
+      error: \`Failed to submit workflow: HTTP \${submitRes.status}\`,
+      detail: submitRes.body,
+    };
+  }
+
+  const promptId = submitRes.body.prompt_id;
+  console.log(\`[comfyui-video] Submitted prompt \${promptId}, waiting for video...\`);
+
+  const result = await pollUntilDone(promptId);
+  if (!result.success) {
+    return { error: result.error };
+  }
+
+  return {
+    success: true,
+    promptId,
+    outputs: result.outputs,
+    message: \`Generated \${result.outputs.length} output(s). URLs:\`,
+    urls: result.outputs.map((o) => o.url),
+  };
+}
+
+comfyuiVideoHandler.execute = async (task, _ctx, _skill) => {
+  const input = typeof task === "string" ? task : (task.input || task.params?.input || "");
+  let p = {};
+  try { p = input.trim().startsWith("{") ? JSON.parse(input) : { prompt: input }; }
+  catch { p = { prompt: input }; }
+  return comfyuiVideoHandler(p);
+};
+module.exports = comfyuiVideoHandler;
+`,
+  },
+  "audio-gen": {
+    md: `---
+name: audio-gen
+display-name: AI 音频生成
+category: media
+description: AI 音频生成（TTS 语音合成 / 音乐生成），支持 edge-tts、piper-tts、ElevenLabs、OpenAI TTS
+version: 1.0.0
+author: ChainlessChain
+parameters:
+  - name: text
+    type: string
+    required: true
+    description: 要合成的文本内容
+  - name: voice
+    type: string
+    required: false
+    description: 语音名称（edge-tts 示例：zh-CN-XiaoxiaoNeural；OpenAI 示例：alloy）
+    default: "zh-CN-XiaoxiaoNeural"
+  - name: type
+    type: string
+    required: false
+    description: 生成类型：tts（语音合成）
+    default: "tts"
+  - name: output
+    type: string
+    required: false
+    description: 输出文件路径（.mp3 或 .wav）
+    default: "output.mp3"
+execution-mode: direct
+---
+
+# AI 音频生成
+
+支持多种后端的 AI 语音合成技能，按以下优先级自动选择可用后端：
+
+1. **edge-tts**（推荐，免费）— 微软 Edge TTS，支持多语言
+   \`\`\`bash
+   pip install edge-tts
+   \`\`\`
+
+2. **piper-tts**（离线，免费）— 高质量神经网络 TTS
+   \`\`\`bash
+   pip install piper-tts
+   \`\`\`
+
+3. **ElevenLabs API**（高质量，付费）— 需设置 \`ELEVENLABS_API_KEY\` 环境变量
+
+4. **OpenAI TTS API**（高质量，付费）— 需设置 \`OPENAI_API_KEY\` 环境变量
+
+## 使用示例
+
+\`\`\`bash
+# 中文语音合成（需安装 edge-tts）
+chainlesschain skill run audio-gen "你好，欢迎使用 ChainlessChain AI 音频生成功能"
+
+# 英文语音，指定输出文件
+chainlesschain skill run audio-gen "Hello world" --args '{"voice":"en-US-AriaNeural","output":"hello.mp3"}'
+
+# 使用 OpenAI TTS
+OPENAI_API_KEY=sk-xxx chainlesschain skill run audio-gen "Hello" --args '{"voice":"alloy"}'
+\`\`\`
+
+## cli-anything 集成
+
+如果有支持 CLI 的音频生成工具（如 audiogen-cli、bark-cli 等），可以通过以下命令注册：
+
+\`\`\`bash
+chainlesschain cli-anything register <tool-name>
+\`\`\`
+`,
+    handler: `/**
+ * AI Audio Generation Skill Handler
+ * Supports multiple TTS backends: edge-tts, piper-tts, ElevenLabs API, OpenAI TTS.
+ * Auto-detects which backend is available and uses the best one.
+ */
+
+const { execSync, spawn } = require("child_process");
+const fs = require("fs");
+const path = require("path");
+const https = require("https");
+
+function commandExists(cmd) {
+  try {
+    execSync(\`\${cmd} --version\`, { stdio: "ignore", encoding: "utf-8" });
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+function runEdgeTTS(text, voice, outputPath) {
+  return new Promise((resolve, reject) => {
+    const args = ["edge-tts", "--voice", voice || "zh-CN-XiaoxiaoNeural", "--text", text, "--write-media", outputPath];
+    const proc = spawn("python", ["-m", ...args], { stdio: ["ignore", "pipe", "pipe"] });
+    let stderr = "";
+    proc.stderr.on("data", (d) => (stderr += d.toString("utf8")));
+    proc.on("close", (code) => {
+      if (code === 0) resolve(outputPath);
+      else reject(new Error(\`edge-tts exited \${code}: \${stderr}\`));
+    });
+  });
+}
+
+function callElevenLabsAPI(text, voice, outputPath, apiKey) {
+  return new Promise((resolve, reject) => {
+    const voiceId = voice || "21m00Tcm4TlvDq8ikWAM"; // Default: Rachel
+    const payload = JSON.stringify({
+      text,
+      model_id: "eleven_monolingual_v1",
+      voice_settings: { stability: 0.5, similarity_boost: 0.5 },
+    });
+    const options = {
+      hostname: "api.elevenlabs.io",
+      path: \`/v1/text-to-speech/\${voiceId}\`,
+      method: "POST",
+      headers: {
+        "xi-api-key": apiKey,
+        "Content-Type": "application/json",
+        Accept: "audio/mpeg",
+      },
+    };
+    const req = https.request(options, (res) => {
+      if (res.statusCode !== 200) {
+        reject(new Error(\`ElevenLabs API returned \${res.statusCode}\`));
+        return;
+      }
+      const out = fs.createWriteStream(outputPath);
+      res.pipe(out);
+      out.on("finish", () => resolve(outputPath));
+    });
+    req.on("error", reject);
+    req.write(payload);
+    req.end();
+  });
+}
+
+function callOpenAITTS(text, voice, outputPath, apiKey) {
+  return new Promise((resolve, reject) => {
+    const payload = JSON.stringify({
+      model: "tts-1",
+      input: text,
+      voice: voice || "alloy",
+    });
+    const options = {
+      hostname: "api.openai.com",
+      path: "/v1/audio/speech",
+      method: "POST",
+      headers: {
+        Authorization: \`Bearer \${apiKey}\`,
+        "Content-Type": "application/json",
+      },
+    };
+    const req = https.request(options, (res) => {
+      if (res.statusCode !== 200) {
+        reject(new Error(\`OpenAI TTS API returned \${res.statusCode}\`));
+        return;
+      }
+      const out = fs.createWriteStream(outputPath);
+      res.pipe(out);
+      out.on("finish", () => resolve(outputPath));
+    });
+    req.on("error", reject);
+    req.write(payload);
+    req.end();
+  });
+}
+
+async function audioGenHandler(params) {
+  const { text, voice, type, output } = params;
+
+  if (!text) {
+    return { error: "Missing required parameter: text" };
+  }
+
+  const outputPath = output || "output.mp3";
+  const outputDir = path.dirname(path.resolve(outputPath));
+  if (!fs.existsSync(outputDir)) {
+    fs.mkdirSync(outputDir, { recursive: true });
+  }
+
+  // Try backends in priority order
+  const elevenLabsKey = process.env.ELEVENLABS_API_KEY;
+  const openaiKey = process.env.OPENAI_API_KEY;
+
+  // 1. edge-tts (free, no API key needed)
+  if (commandExists("edge-tts") || await checkPythonModule("edge_tts")) {
+    try {
+      await runEdgeTTS(text, voice, outputPath);
+      return {
+        success: true,
+        backend: "edge-tts",
+        output: outputPath,
+        message: \`Audio saved to \${outputPath} (edge-tts)\`,
+      };
+    } catch (err) {
+      console.warn("[audio-gen] edge-tts failed:", err.message);
+    }
+  }
+
+  // 2. piper-tts (offline, free)
+  if (commandExists("piper")) {
+    try {
+      execSync(\`echo "\${text.replace(/"/g, '\\\\"')}" | piper --output_file "\${outputPath}"\`, {
+        encoding: "utf-8",
+      });
+      return {
+        success: true,
+        backend: "piper-tts",
+        output: outputPath,
+        message: \`Audio saved to \${outputPath} (piper-tts)\`,
+      };
+    } catch (err) {
+      console.warn("[audio-gen] piper-tts failed:", err.message);
+    }
+  }
+
+  // 3. ElevenLabs API
+  if (elevenLabsKey) {
+    try {
+      await callElevenLabsAPI(text, voice, outputPath, elevenLabsKey);
+      return {
+        success: true,
+        backend: "elevenlabs",
+        output: outputPath,
+        message: \`Audio saved to \${outputPath} (ElevenLabs)\`,
+      };
+    } catch (err) {
+      console.warn("[audio-gen] ElevenLabs failed:", err.message);
+    }
+  }
+
+  // 4. OpenAI TTS
+  if (openaiKey) {
+    try {
+      await callOpenAITTS(text, voice, outputPath, openaiKey);
+      return {
+        success: true,
+        backend: "openai-tts",
+        output: outputPath,
+        message: \`Audio saved to \${outputPath} (OpenAI TTS)\`,
+      };
+    } catch (err) {
+      console.warn("[audio-gen] OpenAI TTS failed:", err.message);
+    }
+  }
+
+  // No backend available
+  return {
+    error: "No TTS backend available",
+    hint: [
+      "Install one of the following:",
+      "  1. edge-tts (free):     pip install edge-tts",
+      "  2. piper-tts (offline): pip install piper-tts",
+      "  3. ElevenLabs (paid):   export ELEVENLABS_API_KEY=<your-key>",
+      "  4. OpenAI TTS (paid):   export OPENAI_API_KEY=<your-key>",
+    ].join("\\n"),
+  };
+};
+
+async function checkPythonModule(moduleName) {
+  try {
+    execSync(\`python -c "import \${moduleName}"\`, { stdio: "ignore", encoding: "utf-8" });
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+audioGenHandler.execute = async (task, _ctx, _skill) => {
+  const input = typeof task === "string" ? task : (task.input || task.params?.input || "");
+  let p = {};
+  try { p = input.trim().startsWith("{") ? JSON.parse(input) : { text: input }; }
+  catch { p = { text: input }; }
+  return audioGenHandler(p);
+};
+module.exports = audioGenHandler;
+`,
+  },
+};
+
+const WORKFLOW_README = `# ComfyUI Workflows
+
+此目录存放 ComfyUI 工作流 JSON 文件。
+
+## 如何导出工作流
+
+1. 在 ComfyUI 界面中构建你的工作流
+2. 点击菜单 → Save (API Format) 保存为 API 格式 JSON
+3. 将文件保存到此目录
+
+## 使用工作流
+
+\`\`\`bash
+# 图像生成
+chainlesschain skill run comfyui-image "a sunset over mountains" --args '{"workflow":"workflows/my-image-workflow.json"}'
+
+# 视频生成（AnimateDiff）
+chainlesschain skill run comfyui-video "a cat walking" --args '{"workflow":"workflows/my-animatediff-workflow.json","frames":16}'
+\`\`\`
+
+## 工作流资源
+
+- ComfyUI 官方示例：https://github.com/comfyanonymous/ComfyUI/tree/master/tests/inference
+- AnimateDiff 工作流：https://github.com/guoyww/AnimateDiff-EvolutionaryFramework
+- ComfyUI Manager（管理扩展）：https://github.com/ltdrdata/ComfyUI-Manager
+
+## cli-anything 集成
+
+如果你安装了有 CLI 接口的 AI 工具，可以通过以下命令注册到 ChainlessChain：
+
+\`\`\`bash
+# 检查可用的 CLI AI 工具
+chainlesschain cli-anything scan
+
+# 注册工具
+chainlesschain cli-anything register <tool-name>
+
+# 查看已注册工具
+chainlesschain cli-anything list
+\`\`\`
+
+适合通过 cli-anything 注册的工具（有 CLI 接口）：
+- FFmpeg（视频处理）
+- yt-dlp（视频下载）
+- audiogen-cli（音频生成 CLI）
+- 第三方 ComfyUI CLI 包装脚本
+`;
+
+// Workspace skill templates for ai-doc-creator
+Object.assign(SKILL_TEMPLATES, {
+  "doc-generate": {
+    md: `---
+name: doc-generate
+display-name: AI 文档生成
+category: document
+description: 利用 AI 生成结构化文档（报告/方案/说明书），支持 Markdown/DOCX/PDF 输出
+version: 1.0.0
+author: ChainlessChain
+parameters:
+  - name: topic
+    type: string
+    required: true
+    description: 文档主题或标题
+  - name: format
+    type: string
+    required: false
+    description: 输出格式：md（默认）/ docx / pdf / html
+    default: "md"
+  - name: outline
+    type: string
+    required: false
+    description: 文档大纲（可选，不提供则 AI 自动规划）
+  - name: style
+    type: string
+    required: false
+    description: 文档风格：report（报告）/ proposal（方案）/ manual（说明书）/ readme（README）
+    default: "report"
+  - name: output
+    type: string
+    required: false
+    description: 输出文件名（不含扩展名，默认为主题名）
+execution-mode: direct
+---
+
+# AI 文档生成
+
+利用 AI 生成结构化文档，支持多种格式输出。
+
+## 使用示例
+
+\`\`\`bash
+# 生成 Markdown 报告
+chainlesschain skill run doc-generate "2026年AI技术趋势分析报告"
+
+# 生成方案文档（指定风格）
+chainlesschain skill run doc-generate "电商平台重构方案" --args '{"style":"proposal","format":"md"}'
+
+# 生成 DOCX（需要 LibreOffice 或 pandoc）
+chainlesschain skill run doc-generate "产品需求说明书" --args '{"format":"docx","style":"manual"}'
+
+# 提供大纲（精准控制结构）
+chainlesschain skill run doc-generate "API文档" --args '{"outline":"1.概述 2.认证方式 3.接口列表 4.错误码","format":"md"}'
+\`\`\`
+
+## 格式转换依赖
+
+| 输出格式 | 依赖 | 安装方式 |
+|---------|------|---------|
+| md | 无 | 内置 |
+| html | 无 | 内置（markdown 渲染） |
+| docx | pandoc 或 LibreOffice | 见下方 |
+| pdf | LibreOffice | \`soffice --headless\` |
+
+\`\`\`bash
+# 安装 pandoc（推荐，md→docx 最佳）
+# Windows: winget install pandoc  或  choco install pandoc
+# macOS:   brew install pandoc
+# Linux:   apt install pandoc
+
+# LibreOffice（全格式转换）
+# Windows: winget install LibreOffice.LibreOffice
+# macOS:   brew install --cask libreoffice
+# Linux:   apt install libreoffice
+\`\`\`
+
+## cli-anything 集成说明
+
+LibreOffice 具有完整的 CLI 接口（\`soffice --headless\`），可以通过 cli-anything 注册以获得更直接的访问方式：
+
+\`\`\`bash
+# 注册 LibreOffice 为独立技能
+chainlesschain cli-anything register soffice
+# 或
+chainlesschain cli-anything register libreoffice
+
+# 注册后可以直接调用 LibreOffice 任意子命令
+chainlesschain skill run cli-anything-soffice "convert report.md to pdf"
+\`\`\`
+
+> 建议：日常 AI 文档生成使用 \`doc-generate\` 技能；需要直接操作 LibreOffice 高级功能（宏、模板、样式）时，通过 cli-anything 注册 \`soffice\` 获得更大灵活性。
+`,
+    handler: `/**
+ * AI Document Generation Skill Handler
+ * Generates structured documents using the local LLM (via chainlesschain ask),
+ * then optionally converts to DOCX/PDF using pandoc or LibreOffice.
+ *
+ * Requirements for format conversion:
+ *   - md/html: built-in (no dependencies)
+ *   - docx:    pandoc (preferred) or LibreOffice soffice --headless
+ *   - pdf:     LibreOffice soffice --headless
+ */
+
+const { spawnSync, execSync } = require("child_process");
+const fs = require("fs");
+const path = require("path");
+
+// ─── Tool detection ───────────────────────────────────────────────
+
+function commandExists(cmd) {
+  try {
+    execSync(\`\${cmd} --version\`, { stdio: "ignore", encoding: "utf-8" });
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+function detectSoffice() {
+  if (commandExists("soffice")) return "soffice";
+  if (commandExists("libreoffice")) return "libreoffice";
+  // Windows default install path
+  const winPath = "C:\\\\Program Files\\\\LibreOffice\\\\program\\\\soffice.exe";
+  if (process.platform === "win32" && fs.existsSync(winPath)) return \`"\${winPath}"\`;
+  return null;
+}
+
+// ─── Document style prompts ───────────────────────────────────────
+
+const STYLE_PROMPTS = {
+  report: "一份专业的分析报告，包含执行摘要、背景、详细分析、结论和建议",
+  proposal: "一份详细的项目方案，包含项目背景、目标、实施方案、资源需求、风险分析和预期收益",
+  manual: "一份完整的说明书/手册，包含概述、安装/配置步骤、功能说明、常见问题和参考资料",
+  readme: "一份标准的 README 文档，包含项目简介、快速开始、功能特性、安装使用、贡献指南和许可证",
+};
+
+// ─── Markdown to HTML ─────────────────────────────────────────────
+
+function mdToHtml(md, title) {
+  return \`<!DOCTYPE html>
+<html lang="zh-CN">
+<head>
+  <meta charset="UTF-8">
+  <meta name="viewport" content="width=device-width, initial-scale=1.0">
+  <title>\${title}</title>
+  <style>
+    body { font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif;
+           max-width: 900px; margin: 0 auto; padding: 40px 20px; line-height: 1.6; }
+    h1, h2, h3 { border-bottom: 1px solid #eee; padding-bottom: 8px; }
+    pre { background: #f6f8fa; padding: 16px; border-radius: 6px; overflow: auto; }
+    code { background: #f0f0f0; padding: 2px 6px; border-radius: 3px; }
+    table { border-collapse: collapse; width: 100%; }
+    th, td { border: 1px solid #ddd; padding: 8px 12px; text-align: left; }
+    th { background: #f6f8fa; }
+  </style>
+</head>
+<body>
+\${md
+  .replace(/^### (.+)$/gm, "<h3>$1</h3>")
+  .replace(/^## (.+)$/gm, "<h2>$1</h2>")
+  .replace(/^# (.+)$/gm, "<h1>$1</h1>")
+  .replace(/\\*\\*(.+?)\\*\\*/g, "<strong>$1</strong>")
+  .replace(/\\*(.+?)\\*/g, "<em>$1</em>")
+  .replace(/\`(.+?)\`/g, "<code>$1</code>")
+  .replace(/^- (.+)$/gm, "<li>$1</li>")
+  .replace(/(<li>.*<\\/li>\\n?)+/g, (m) => "<ul>" + m + "</ul>")
+  .replace(/^\\d+\\. (.+)$/gm, "<li>$1</li>")
+  .replace(/\\n\\n/g, "</p><p>")
+  .replace(/^(?!<[h|u|o|l|p|t])/gm, "<p>")
+}
+</body>
+</html>\`;
+}
+
+// ─── Conversion helpers ───────────────────────────────────────────
+
+function convertWithPandoc(mdFile, outputFile, format) {
+  const result = spawnSync(
+    "pandoc",
+    [mdFile, "-o", outputFile, "--standalone"],
+    { encoding: "utf-8", timeout: 60000 },
+  );
+  if (result.status !== 0) {
+    throw new Error(\`pandoc failed: \${result.stderr || result.error?.message}\`);
+  }
+  return outputFile;
+}
+
+function convertWithSoffice(sofficeCmd, inputFile, format, outDir) {
+  const result = spawnSync(
+    sofficeCmd,
+    ["--headless", "--convert-to", format, inputFile, "--outdir", outDir],
+    { encoding: "utf-8", timeout: 120000, shell: process.platform === "win32" },
+  );
+  if (result.status !== 0) {
+    throw new Error(\`soffice failed: \${result.stderr || result.error?.message}\`);
+  }
+  // soffice outputs to <basename>.<format> in outDir
+  const base = path.basename(inputFile, path.extname(inputFile));
+  return path.join(outDir, \`\${base}.\${format}\`);
+}
+
+// ─── LLM content generation ───────────────────────────────────────
+
+function generateContent(topic, style, outline) {
+  const styleDesc = STYLE_PROMPTS[style] || STYLE_PROMPTS.report;
+  const outlineSection = outline
+    ? \`\\n\\n大纲要求（请按以下结构组织内容）：\\n\${outline}\`
+    : "";
+
+  const prompt =
+    \`请为以下主题生成\${styleDesc}。\\n\\n主题：\${topic}\${outlineSection}\\n\\n要求：\\n- 使用标准 Markdown 格式（H1/H2/H3 标题、列表、表格等）\\n- 内容详实具体，不少于 800 字\\n- 使用中文撰写\\n- 直接输出 Markdown 内容，不需要额外说明\`;
+
+  // Try chainlesschain ask first
+  const askResult = spawnSync(
+    process.execPath,
+    [process.argv[1], "ask", prompt],
+    {
+      encoding: "utf-8",
+      timeout: 120000,
+      cwd: process.cwd(),
+      env: process.env,
+    },
+  );
+
+  if (askResult.status === 0 && askResult.stdout && askResult.stdout.trim().length > 100) {
+    return askResult.stdout.trim();
+  }
+
+  // Fallback: return a structured template that user can fill
+  return \`# \${topic}
+
+> *本文档由 AI 文档生成技能创建。LLM 调用失败，请手动完善内容，或确认 chainlesschain ask 命令可用。*
+
+## 概述
+
+[在此填写 \${topic} 的概述内容]
+
+## 背景与目标
+
+[在此填写背景信息和文档目标]
+
+## 详细内容
+
+### 一、[主要章节]
+
+[在此填写主要内容]
+
+### 二、[次要章节]
+
+[在此填写次要内容]
+
+## 结论与建议
+
+[在此填写结论]
+
+## 附录
+
+[在此填写附录内容（如有）]
+\`;
+}
+
+// ─── Main handler ─────────────────────────────────────────────────
+
+async function docGenerateHandler(params) {
+  const { topic, format, outline, style, output } = params;
+
+  if (!topic) {
+    return { error: "Missing required parameter: topic", hint: "chainlesschain skill run doc-generate \\"文档主题\\"" };
+  }
+
+  const fmt = (format || "md").toLowerCase();
+  const docStyle = style || "report";
+  const safeTitle = (output || topic).replace(/[^a-zA-Z0-9\u4e00-\u9fa5_-]/g, "_").slice(0, 50);
+  const outputDir = process.cwd();
+
+  // Generate content via LLM
+  console.log(\`[doc-generate] Generating \${docStyle} document for: \${topic}\`);
+  const mdContent = generateContent(topic, docStyle, outline);
+
+  // Always write markdown first
+  const mdFile = path.join(outputDir, \`\${safeTitle}.md\`);
+  fs.writeFileSync(mdFile, mdContent, "utf-8");
+  console.log(\`[doc-generate] Markdown saved: \${mdFile}\`);
+
+  if (fmt === "md") {
+    return {
+      success: true,
+      format: "md",
+      output: mdFile,
+      message: \`Markdown document saved to: \${mdFile}\`,
+      wordCount: mdContent.split(/\\s+/).length,
+    };
+  }
+
+  if (fmt === "html") {
+    const htmlFile = path.join(outputDir, \`\${safeTitle}.html\`);
+    fs.writeFileSync(htmlFile, mdToHtml(mdContent, topic), "utf-8");
+    return {
+      success: true,
+      format: "html",
+      output: htmlFile,
+      mdOutput: mdFile,
+      message: \`HTML document saved to: \${htmlFile}\`,
+    };
+  }
+
+  // DOCX / PDF: try pandoc then soffice
+  const sofficeCmd = detectSoffice();
+
+  if (fmt === "docx") {
+    if (commandExists("pandoc")) {
+      try {
+        const docxFile = path.join(outputDir, \`\${safeTitle}.docx\`);
+        convertWithPandoc(mdFile, docxFile, "docx");
+        return { success: true, format: "docx", output: docxFile, mdOutput: mdFile, message: \`DOCX saved to: \${docxFile} (via pandoc)\` };
+      } catch (err) {
+        console.warn("[doc-generate] pandoc failed:", err.message);
+      }
+    }
+    if (sofficeCmd) {
+      try {
+        // soffice needs HTML or ODT as input for md→docx (md not natively supported)
+        const htmlFile = path.join(outputDir, \`\${safeTitle}.html\`);
+        fs.writeFileSync(htmlFile, mdToHtml(mdContent, topic), "utf-8");
+        const docxFile = convertWithSoffice(sofficeCmd, htmlFile, "docx", outputDir);
+        // Clean up temp HTML
+        try { fs.unlinkSync(htmlFile); } catch { /* ignore */ }
+        return { success: true, format: "docx", output: docxFile, mdOutput: mdFile, message: \`DOCX saved to: \${docxFile} (via LibreOffice)\` };
+      } catch (err) {
+        console.warn("[doc-generate] soffice docx failed:", err.message);
+      }
+    }
+    return {
+      success: true,
+      format: "md",
+      output: mdFile,
+      message: \`Markdown saved (DOCX conversion requires pandoc or LibreOffice — neither found)\`,
+      hint: "Install: winget install pandoc  OR  apt install pandoc  OR  apt install libreoffice",
+    };
+  }
+
+  if (fmt === "pdf") {
+    if (sofficeCmd) {
+      try {
+        const htmlFile = path.join(outputDir, \`\${safeTitle}.html\`);
+        fs.writeFileSync(htmlFile, mdToHtml(mdContent, topic), "utf-8");
+        const pdfFile = convertWithSoffice(sofficeCmd, htmlFile, "pdf", outputDir);
+        try { fs.unlinkSync(htmlFile); } catch { /* ignore */ }
+        return { success: true, format: "pdf", output: pdfFile, mdOutput: mdFile, message: \`PDF saved to: \${pdfFile} (via LibreOffice)\` };
+      } catch (err) {
+        console.warn("[doc-generate] soffice pdf failed:", err.message);
+      }
+    }
+    if (commandExists("pandoc") && commandExists("wkhtmltopdf")) {
+      try {
+        const pdfFile = path.join(outputDir, \`\${safeTitle}.pdf\`);
+        convertWithPandoc(mdFile, pdfFile, "pdf");
+        return { success: true, format: "pdf", output: pdfFile, mdOutput: mdFile, message: \`PDF saved to: \${pdfFile} (via pandoc+wkhtmltopdf)\` };
+      } catch (err) {
+        console.warn("[doc-generate] pandoc pdf failed:", err.message);
+      }
+    }
+    return {
+      success: true,
+      format: "md",
+      output: mdFile,
+      message: \`Markdown saved (PDF conversion requires LibreOffice — not found)\`,
+      hint: "Install: apt install libreoffice  OR  winget install LibreOffice.LibreOffice",
+    };
+  }
+
+  return { error: \`Unsupported format: \${fmt}\`, hint: "Supported formats: md, html, docx, pdf" };
+}
+
+docGenerateHandler.execute = async (task, _ctx, _skill) => {
+  const input = typeof task === "string" ? task : (task.input || task.params?.input || "");
+  let p = {};
+  try { p = input.trim().startsWith("{") ? JSON.parse(input) : { topic: input }; }
+  catch { p = { topic: input }; }
+  return docGenerateHandler(p);
+};
+module.exports = docGenerateHandler;
+`,
+  },
+
+  "libre-convert": {
+    md: `---
+name: libre-convert
+display-name: LibreOffice 文档转换
+category: document
+description: 使用 LibreOffice（soffice --headless）在 docx/pdf/html/odt/pptx 等格式之间转换文档
+version: 1.0.0
+author: ChainlessChain
+parameters:
+  - name: input_file
+    type: string
+    required: true
+    description: 要转换的源文件路径
+  - name: format
+    type: string
+    required: false
+    description: 目标格式：pdf / docx / html / odt / pptx / xlsx / png
+    default: "pdf"
+  - name: outdir
+    type: string
+    required: false
+    description: 输出目录（默认与源文件相同目录）
+execution-mode: direct
+---
+
+# LibreOffice 文档格式转换
+
+使用 LibreOffice 的无头模式（\`soffice --headless\`）在各种办公文档格式之间转换。
+
+## 支持的转换格式
+
+| 源格式 | → 目标格式 |
+|--------|-----------|
+| docx / doc | pdf, odt, html |
+| odt | pdf, docx, html |
+| pptx / ppt | pdf, html, png |
+| xlsx / xls | pdf, csv, html |
+| html / md | pdf, docx, odt |
+| jpg / png | pdf |
+
+## 使用示例
+
+\`\`\`bash
+# Word 文档转 PDF
+chainlesschain skill run libre-convert "report.docx"
+
+# 指定格式
+chainlesschain skill run libre-convert "slides.pptx" --args '{"format":"pdf"}'
+
+# 指定输出目录
+chainlesschain skill run libre-convert "contract.docx" --args '{"format":"pdf","outdir":"./output"}'
+
+# 批量转换（在 agent 模式下）
+chainlesschain agent
+# > 将 docs/ 目录下所有 docx 文件转换为 PDF
+\`\`\`
+
+## cli-anything 集成
+
+LibreOffice 具有完整的 CLI 接口，除了使用此技能外，还可以通过 cli-anything 将完整的 \`soffice\` 命令行注册为技能：
+
+\`\`\`bash
+# 注册完整的 soffice CLI（支持 LibreOffice 所有子命令）
+chainlesschain cli-anything register soffice
+
+# 注册后可访问 LibreOffice 所有功能（宏执行、模板、高级格式设置等）
+chainlesschain skill run cli-anything-soffice "convert report.docx --format pdf"
+\`\`\`
+
+## 前提条件
+
+需要安装 LibreOffice：
+
+\`\`\`bash
+# Windows
+winget install LibreOffice.LibreOffice
+
+# macOS
+brew install --cask libreoffice
+
+# Ubuntu/Debian
+sudo apt install libreoffice
+
+# 验证安装
+soffice --version
+\`\`\`
+`,
+    handler: `/**
+ * LibreOffice Document Conversion Skill Handler
+ * Uses soffice --headless to convert between document formats.
+ *
+ * Requirements: LibreOffice installed (soffice or libreoffice in PATH,
+ * or Windows default installation path)
+ *
+ * LibreOffice CLI vs cli-anything:
+ *   - This skill: embedded wrapper for common conversion use cases
+ *   - cli-anything: register soffice for full LibreOffice CLI access
+ *     chainlesschain cli-anything register soffice
+ */
+
+const { spawnSync, execSync } = require("child_process");
+const fs = require("fs");
+const path = require("path");
+
+const SUPPORTED_FORMATS = new Set(["pdf", "docx", "html", "odt", "pptx", "xlsx", "csv", "txt", "png"]);
+
+function findSoffice() {
+  const candidates = ["soffice", "libreoffice"];
+  for (const cmd of candidates) {
+    try {
+      execSync(\`\${cmd} --version\`, { stdio: "ignore", encoding: "utf-8" });
+      return cmd;
+    } catch { /* try next */ }
+  }
+  // Windows default installation paths
+  if (process.platform === "win32") {
+    const paths = [
+      "C:\\\\Program Files\\\\LibreOffice\\\\program\\\\soffice.exe",
+      "C:\\\\Program Files (x86)\\\\LibreOffice\\\\program\\\\soffice.exe",
+    ];
+    for (const p of paths) {
+      if (fs.existsSync(p)) return \`"\${p}"\`;
+    }
+  }
+  return null;
+}
+
+async function libreConvertHandler(params) {
+  const { input_file, format, outdir } = params;
+
+  if (!input_file) {
+    return { error: "Missing required parameter: input_file" };
+  }
+
+  const targetFormat = (format || "pdf").toLowerCase();
+  if (!SUPPORTED_FORMATS.has(targetFormat)) {
+    return {
+      error: \`Unsupported format: \${targetFormat}\`,
+      hint: \`Supported formats: \${[...SUPPORTED_FORMATS].join(", ")}\`,
+    };
+  }
+
+  // Resolve input file path
+  const inputPath = path.isAbsolute(input_file)
+    ? input_file
+    : path.join(process.cwd(), input_file);
+
+  if (!fs.existsSync(inputPath)) {
+    return { error: \`Input file not found: \${inputPath}\` };
+  }
+
+  // Find LibreOffice
+  const sofficeCmd = findSoffice();
+  if (!sofficeCmd) {
+    return {
+      error: "LibreOffice not found",
+      hint: [
+        "Install LibreOffice to use this skill:",
+        "  Windows: winget install LibreOffice.LibreOffice",
+        "  macOS:   brew install --cask libreoffice",
+        "  Linux:   sudo apt install libreoffice",
+        "",
+        "Alternatively, register soffice via cli-anything for full access:",
+        "  chainlesschain cli-anything register soffice",
+      ].join("\\n"),
+    };
+  }
+
+  // Determine output directory
+  const outputDir = outdir
+    ? (path.isAbsolute(outdir) ? outdir : path.join(process.cwd(), outdir))
+    : path.dirname(inputPath);
+
+  if (!fs.existsSync(outputDir)) {
+    fs.mkdirSync(outputDir, { recursive: true });
+  }
+
+  console.log(\`[libre-convert] Converting \${path.basename(inputPath)} → \${targetFormat}\`);
+
+  // Run soffice
+  const result = spawnSync(
+    sofficeCmd,
+    ["--headless", "--convert-to", targetFormat, inputPath, "--outdir", outputDir],
+    {
+      encoding: "utf-8",
+      timeout: 120000,
+      shell: process.platform === "win32",
+    },
+  );
+
+  if (result.error) {
+    return { error: \`Failed to launch LibreOffice: \${result.error.message}\` };
+  }
+
+  if (result.status !== 0) {
+    return {
+      error: \`LibreOffice conversion failed (exit \${result.status})\`,
+      stderr: result.stderr,
+      hint: "Check if the input file is valid and not password-protected",
+    };
+  }
+
+  // Find the output file
+  const baseName = path.basename(inputPath, path.extname(inputPath));
+  const outputFile = path.join(outputDir, \`\${baseName}.\${targetFormat}\`);
+
+  if (!fs.existsSync(outputFile)) {
+    // soffice may use a slightly different output name — scan dir
+    const files = fs.readdirSync(outputDir).filter(
+      (f) => f.startsWith(baseName) && f.endsWith(\`.\${targetFormat}\`),
+    );
+    if (files.length > 0) {
+      return {
+        success: true,
+        output: path.join(outputDir, files[0]),
+        message: \`Converted to \${targetFormat}: \${files[0]}\`,
+      };
+    }
+    return {
+      error: "Conversion completed but output file not found",
+      stdout: result.stdout,
+      hint: \`Check \${outputDir} for the converted file\`,
+    };
+  }
+
+  return {
+    success: true,
+    input: inputPath,
+    output: outputFile,
+    format: targetFormat,
+    message: \`Converted to \${targetFormat}: \${outputFile}\`,
+  };
+}
+
+libreConvertHandler.execute = async (task, _ctx, _skill) => {
+  const input = typeof task === "string" ? task : (task.input || task.params?.input || "");
+  let p = {};
+  try { p = input.trim().startsWith("{") ? JSON.parse(input) : { input_file: input }; }
+  catch { p = { input_file: input }; }
+  return libreConvertHandler(p);
+};
+module.exports = libreConvertHandler;
+`,
+  },
+});
+
+// Workspace skill templates for doc-edit (ai-doc-creator extension)
+Object.assign(SKILL_TEMPLATES, {
+  "doc-edit": {
+    md: `---
+name: doc-edit
+display-name: AI 文档修改
+category: document
+description: 使用 AI 修改现有文档内容，保留原始格式与结构（公式/图表/样式不丢失）
+version: 1.0.0
+author: ChainlessChain
+input_schema:
+  type: object
+  properties:
+    input_file:
+      type: string
+      description: 要修改的文件路径（支持 md/txt/html/docx/xlsx/pptx）
+    instruction:
+      type: string
+      description: 修改指令，如"将所有'项目'替换为'工程'并优化摘要部分"
+    action:
+      type: string
+      enum: [edit, append, rewrite-section]
+      default: edit
+      description: 操作模式（edit=全文修改, append=追加章节, rewrite-section=重写指定标题）
+    section:
+      type: string
+      description: rewrite-section 时指定目标标题名称
+    output_dir:
+      type: string
+      description: 输出目录（默认与输入文件同目录）
+  required: [input_file, instruction]
+execution-mode: direct
+---
+
+# AI 文档修改 (doc-edit)
+
+使用 AI 对现有文档进行智能修改，**保留原始格式与结构**（公式、图表、样式、动画不丢失）。
+
+## 支持格式
+
+| 格式 | 处理方式 | 结构保留 |
+|------|----------|----------|
+| md / txt / html | 直接文本修改 | 完全保留 |
+| docx | pandoc 往返转换（优先）/ soffice 回退 | 基本保留 |
+| xlsx | Python + openpyxl（保留公式/图表） | 公式以字符串保存，完全保留 |
+| pptx | Python + python-pptx（只改文本 run） | 图表/图片/动画完全不碰 |
+
+## 使用示例
+
+\`\`\`bash
+# 修改 Markdown 文档
+chainlesschain skill run doc-edit --args '{"input_file":"report.md","instruction":"优化摘要部分，使语气更正式"}'
+
+# 追加新章节
+chainlesschain skill run doc-edit --args '{"input_file":"report.md","instruction":"添加结论章节，总结主要发现","action":"append"}'
+
+# 重写特定标题
+chainlesschain skill run doc-edit --args '{"input_file":"spec.md","instruction":"重写为更技术性的描述","action":"rewrite-section","section":"架构设计"}'
+
+# 修改 Excel（保留公式）
+chainlesschain skill run doc-edit --args '{"input_file":"data.xlsx","instruction":"将所有产品名称首字母大写"}'
+
+# 修改 PowerPoint（保留图表/动画）
+chainlesschain skill run doc-edit --args '{"input_file":"slides.pptx","instruction":"将语气改为更正式的商业风格"}'
+\`\`\`
+
+## 输出命名规则
+
+输出文件命名为 \`{原始文件名}_edited.{扩展名}\`，**永不覆盖原文件**。
+
+## 前提条件
+
+- **md/txt/html**：无需额外工具
+- **docx**：pandoc（\`winget install pandoc\`）或 LibreOffice（\`winget install LibreOffice.LibreOffice\`）
+- **xlsx**：Python 3.x + openpyxl（\`pip install openpyxl\`）
+- **pptx**：Python 3.x + python-pptx（\`pip install python-pptx\`）
+
+## cli-anything 集成
+
+本技能通过内联 Python 子进程处理 xlsx/pptx，无需注册 cli-anything。
+对于需要更精细控制的场景（如 VBA 宏、复杂图表编辑），可注册 soffice：
+
+\`\`\`bash
+chainlesschain cli-anything register soffice
+\`\`\`
+`,
+    handler: `/**
+ * doc-edit Skill Handler
+ * Edits existing documents using AI, preserving formulas/charts/styles.
+ *
+ * Supported formats:
+ *   md/txt/html  — direct text edit via LLM
+ *   docx         — pandoc round-trip (fallback: soffice)
+ *   xlsx         — Python + openpyxl (preserves formulas, data_only=False)
+ *   pptx         — Python + python-pptx (only edits text runs, skips charts/images)
+ *
+ * Output: {baseName}_edited.{ext} — never overwrites the original.
+ */
+
+"use strict";
+const fs = require("fs");
+const path = require("path");
+const os = require("os");
+const { execSync, spawnSync } = require("child_process");
+
+// ── Python detection (mirrors cli-anything-bridge pattern) ───────────────────
+function detectPython() {
+  const candidates = ["python", "python3", "py"];
+  for (const cmd of candidates) {
+    try {
+      const r = spawnSync(cmd, ["--version"], { encoding: "utf-8", timeout: 5000 });
+      if (r.status === 0) return { found: true, command: cmd };
+    } catch (_e) { /* try next */ }
+  }
+  return { found: false, command: null };
+}
+
+function checkPythonModule(pyCmd, moduleName) {
+  try {
+    execSync(\`\${pyCmd} -c "import \${moduleName}"\`, { stdio: "ignore", timeout: 10000 });
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+// ── LLM call via CLI sub-process ─────────────────────────────────────────────
+function callLLM(prompt) {
+  const r = spawnSync(
+    process.execPath,
+    [process.argv[1], "ask", prompt],
+    { encoding: "utf-8", timeout: 120000, cwd: process.cwd(), env: process.env },
+  );
+  if (r.status === 0 && r.stdout && r.stdout.trim().length > 10) return r.stdout.trim();
+  return null;
+}
+
+// ── Output path helper ────────────────────────────────────────────────────────
+function buildOutputPath(inputFile, outputDir) {
+  const ext = path.extname(inputFile);
+  const baseName = path.basename(inputFile, ext);
+  const dir = outputDir || path.dirname(inputFile);
+  if (!fs.existsSync(dir)) fs.mkdirSync(dir, { recursive: true });
+  return path.join(dir, \`\${baseName}_edited\${ext}\`);
+}
+
+// ── Prompt builders ───────────────────────────────────────────────────────────
+function buildTextPrompt(content, instruction, action, section) {
+  if (action === "append") {
+    return \`请按照以下指令，在文档末尾追加新内容。只输出追加的新内容，不重复原文档：
+
+指令：\${instruction}
+
+原文档内容：
+\${content}\`;
+  }
+  if (action === "rewrite-section" && section) {
+    return \`请按照以下指令，重写文档中标题为"\${section}"的章节。只输出该章节的完整新内容（包含标题行）：
+
+指令：\${instruction}
+
+原文档内容：
+\${content}\`;
+  }
+  // default: edit
+  return \`请按照以下修改指令修改文档，保持原有结构和格式风格，直接输出修改后的完整文档内容：
+
+修改指令：\${instruction}
+
+原文档内容：
+\${content}\`;
+}
+
+// ── md / txt / html handler ───────────────────────────────────────────────────
+function editText(inputFile, instruction, action, section, outputDir) {
+  const content = fs.readFileSync(inputFile, "utf-8");
+  const prompt = buildTextPrompt(content, instruction, action, section);
+  const llmResult = callLLM(prompt);
+
+  let newContent;
+  if (llmResult) {
+    if (action === "append") {
+      newContent = content + "\\n\\n" + llmResult;
+    } else if (action === "rewrite-section" && section) {
+      const lines = content.split("\\n");
+      const sectionStart = lines.findIndex((l) => /^#{1,6}\s/.test(l) && l.includes(section));
+      if (sectionStart >= 0) {
+        const nextSection = lines.findIndex((l, i) => i > sectionStart && /^#{1,6}\s/.test(l));
+        const before = lines.slice(0, sectionStart).join("\\n");
+        const after = nextSection >= 0 ? lines.slice(nextSection).join("\\n") : "";
+        newContent = (before ? before + "\\n" : "") + llmResult + (after ? "\\n" + after : "");
+      } else {
+        newContent = content + "\\n\\n" + llmResult;
+      }
+    } else {
+      newContent = llmResult;
+    }
+  } else {
+    return { success: false, error: "LLM 调用失败，请检查 chainlesschain ask 是否可用", hint: "运行 chainlesschain llm test 检查 LLM 连接" };
+  }
+
+  const outputFile = buildOutputPath(inputFile, outputDir);
+  fs.writeFileSync(outputFile, newContent, "utf-8");
+  return {
+    success: true,
+    input: inputFile,
+    output: outputFile,
+    action,
+    message: \`Document edited and saved to: \${outputFile}\`,
+  };
+}
+
+// ── docx handler ──────────────────────────────────────────────────────────────
+function editDocx(inputFile, instruction, action, section, outputDir) {
+  // Try pandoc: docx → markdown → LLM → markdown → docx
+  try {
+    const r = spawnSync("pandoc", ["--version"], { encoding: "utf-8", timeout: 5000 });
+    if (r.status === 0) {
+      const tmpMd = path.join(os.tmpdir(), \`doc_edit_\${Date.now()}.md\`);
+      spawnSync("pandoc", [inputFile, "-o", tmpMd], { encoding: "utf-8", timeout: 60000 });
+      if (fs.existsSync(tmpMd)) {
+        const content = fs.readFileSync(tmpMd, "utf-8");
+        const prompt = buildTextPrompt(content, instruction, action, section);
+        const llmResult = callLLM(prompt);
+        if (llmResult) {
+          let newContent = action === "append" ? content + "\\n\\n" + llmResult : llmResult;
+          fs.writeFileSync(tmpMd, newContent, "utf-8");
+          const outputFile = buildOutputPath(inputFile, outputDir);
+          spawnSync("pandoc", [tmpMd, "-o", outputFile], { encoding: "utf-8", timeout: 60000 });
+          try { fs.unlinkSync(tmpMd); } catch (_e) { /* ignore */ }
+          if (fs.existsSync(outputFile)) {
+            return { success: true, input: inputFile, output: outputFile, action, message: \`DOCX edited via pandoc: \${outputFile}\` };
+          }
+        }
+        try { fs.unlinkSync(tmpMd); } catch (_e) { /* ignore */ }
+      }
+    }
+  } catch (_e) { /* pandoc not available, try soffice */ }
+
+  // Try soffice: docx → html → LLM → html → docx
+  const sofficeCandidates = ["soffice", "libreoffice"];
+  if (process.platform === "win32") {
+    const winPaths = [
+      "C:\\\\Program Files\\\\LibreOffice\\\\program\\\\soffice.exe",
+      "C:\\\\Program Files (x86)\\\\LibreOffice\\\\program\\\\soffice.exe",
+    ];
+    for (const p of winPaths) {
+      if (fs.existsSync(p)) sofficeCandidates.unshift(p);
+    }
+  }
+  for (const soffice of sofficeCandidates) {
+    try {
+      const rv = spawnSync(soffice, ["--version"], { encoding: "utf-8", timeout: 5000, shell: process.platform === "win32" });
+      if (rv.status === 0) {
+        const tmpDir2 = os.tmpdir();
+        spawnSync(soffice, ["--headless", "--convert-to", "html", inputFile, "--outdir", tmpDir2], {
+          encoding: "utf-8", timeout: 60000, shell: process.platform === "win32",
+        });
+        const baseName = path.basename(inputFile, ".docx");
+        const htmlFile = path.join(tmpDir2, \`\${baseName}.html\`);
+        if (fs.existsSync(htmlFile)) {
+          const content = fs.readFileSync(htmlFile, "utf-8");
+          const prompt = buildTextPrompt(content, instruction, action, section);
+          const llmResult = callLLM(prompt);
+          if (llmResult) {
+            fs.writeFileSync(htmlFile, llmResult, "utf-8");
+            const outputFile = buildOutputPath(inputFile, outputDir);
+            spawnSync(soffice, ["--headless", "--convert-to", "docx", htmlFile, "--outdir", path.dirname(outputFile)], {
+              encoding: "utf-8", timeout: 60000, shell: process.platform === "win32",
+            });
+            try { fs.unlinkSync(htmlFile); } catch (_e) { /* ignore */ }
+            if (fs.existsSync(outputFile)) {
+              return { success: true, input: inputFile, output: outputFile, action, message: \`DOCX edited via LibreOffice: \${outputFile}\` };
+            }
+          }
+          try { fs.unlinkSync(htmlFile); } catch (_e) { /* ignore */ }
+        }
+        break;
+      }
+    } catch (_e) { /* try next */ }
+  }
+
+  return {
+    success: false,
+    error: "需要安装 pandoc 或 LibreOffice 才能修改 DOCX 文件",
+    hint: "安装 pandoc: winget install pandoc  或  安装 LibreOffice: winget install LibreOffice.LibreOffice",
+  };
+}
+
+// ── xlsx handler (Python + openpyxl) ─────────────────────────────────────────
+function editXlsx(inputFile, instruction, action, outputDir) {
+  const py = detectPython();
+  if (!py.found) {
+    return { success: false, error: "未找到 Python，请安装 Python 3.x", hint: "https://python.org" };
+  }
+  if (!checkPythonModule(py.command, "openpyxl")) {
+    return {
+      success: false,
+      error: "需要安装 openpyxl 才能修改 XLSX 文件",
+      hint: \`运行: \${py.command} -m pip install openpyxl\`,
+    };
+  }
+
+  const tmpExtract = path.join(os.tmpdir(), \`xlsx_extract_\${Date.now()}.py\`);
+  const tmpApply = path.join(os.tmpdir(), \`xlsx_apply_\${Date.now()}.py\`);
+  const tmpJson = path.join(os.tmpdir(), \`xlsx_cells_\${Date.now()}.json\`);
+
+  try {
+    // Step 1: extract text cells (non-formula string cells)
+    const extractScript = \`
+import json, openpyxl
+wb = openpyxl.load_workbook(r"""\${inputFile}""", data_only=False)
+cells = []
+for ws in wb.worksheets:
+    for row in ws.iter_rows():
+        for cell in row:
+            if cell.data_type == 's' and cell.value is not None:
+                cells.append({"sheet": ws.title, "row": cell.row, "col": cell.column, "value": cell.value})
+with open(r"""\${tmpJson}""", "w", encoding="utf-8") as f:
+    json.dump(cells, f, ensure_ascii=False)
+\`;
+    fs.writeFileSync(tmpExtract, extractScript, "utf-8");
+    const er = spawnSync(py.command, [tmpExtract], { encoding: "utf-8", timeout: 30000 });
+    if (er.status !== 0) throw new Error(er.stderr || "extract failed");
+
+    const cells = JSON.parse(fs.readFileSync(tmpJson, "utf-8"));
+    if (cells.length === 0) {
+      return { success: false, error: "未找到可修改的文本单元格（所有单元格均为公式或数值）" };
+    }
+
+    // Step 2: LLM editing
+    const textList = cells.map((c, i) => \`[\${i}] \${c.sheet}!R\${c.row}C\${c.col}: \${c.value}\`).join("\\n");
+    const prompt = \`请按照以下修改指令修改 Excel 文本单元格内容。只修改需要修改的单元格，不需要修改的保持原值。
+以 JSON 数组格式返回所有单元格（包括未修改的），每个元素保持原有的 sheet/row/col 字段，只更新 value 字段。
+
+修改指令：\${instruction}
+
+当前文本单元格：
+\${textList}
+
+直接返回 JSON 数组，不要其他说明。\`;
+    const llmResult = callLLM(prompt);
+    if (!llmResult) {
+      return { success: false, error: "LLM 调用失败", hint: "运行 chainlesschain llm test 检查连接" };
+    }
+
+    // Parse LLM result
+    let updatedCells;
+    try {
+      const jsonMatch = llmResult.match(/\\[\\s*\\{[\\s\\S]*\\}\\s*\\]/);
+      updatedCells = JSON.parse(jsonMatch ? jsonMatch[0] : llmResult);
+    } catch (_e) {
+      return { success: false, error: "LLM 返回格式不正确，无法解析 JSON", hint: "请重试或简化修改指令" };
+    }
+
+    // Step 3: apply changes
+    const outputFile = buildOutputPath(inputFile, outputDir);
+    const applyScript = \`
+import json, openpyxl, shutil
+shutil.copy2(r"""\${inputFile}""", r"""\${outputFile}""")
+wb = openpyxl.load_workbook(r"""\${outputFile}""", data_only=False)
+updated = json.loads('''
+\${JSON.stringify(updatedCells).replace(/'/g, "\\\\'")}
+''')
+for item in updated:
+    ws = wb[item["sheet"]]
+    cell = ws.cell(row=item["row"], column=item["col"])
+    if cell.data_type == 's':
+        cell.value = item["value"]
+wb.save(r"""\${outputFile}""")
+\`;
+    fs.writeFileSync(tmpApply, applyScript, "utf-8");
+    const ar = spawnSync(py.command, [tmpApply], { encoding: "utf-8", timeout: 30000 });
+    if (ar.status !== 0) throw new Error(ar.stderr || "apply failed");
+
+    return {
+      success: true,
+      input: inputFile,
+      output: outputFile,
+      action: action || "edit",
+      message: \`XLSX edited (formulas preserved): \${outputFile}\`,
+      cellsModified: updatedCells.length,
+    };
+  } catch (err) {
+    return { success: false, error: \`XLSX 修改失败: \${err.message}\` };
+  } finally {
+    try { fs.unlinkSync(tmpExtract); } catch (_e) { /* ignore */ }
+    try { fs.unlinkSync(tmpApply); } catch (_e) { /* ignore */ }
+    try { fs.unlinkSync(tmpJson); } catch (_e) { /* ignore */ }
+  }
+}
+
+// ── pptx handler (Python + python-pptx) ──────────────────────────────────────
+function editPptx(inputFile, instruction, action, outputDir) {
+  const py = detectPython();
+  if (!py.found) {
+    return { success: false, error: "未找到 Python，请安装 Python 3.x", hint: "https://python.org" };
+  }
+  if (!checkPythonModule(py.command, "pptx")) {
+    return {
+      success: false,
+      error: "需要安装 python-pptx 才能修改 PPTX 文件",
+      hint: \`运行: \${py.command} -m pip install python-pptx\`,
+    };
+  }
+
+  const tmpExtract = path.join(os.tmpdir(), \`pptx_extract_\${Date.now()}.py\`);
+  const tmpApply = path.join(os.tmpdir(), \`pptx_apply_\${Date.now()}.py\`);
+  const tmpJson = path.join(os.tmpdir(), \`pptx_runs_\${Date.now()}.json\`);
+
+  try {
+    // Step 1: extract text runs (skip chart shapes)
+    const extractScript = \`
+import json
+from pptx import Presentation
+prs = Presentation(r"""\${inputFile}""")
+runs = []
+for si, slide in enumerate(prs.slides):
+    for shi, shape in enumerate(slide.shapes):
+        if not shape.has_text_frame:
+            continue
+        for pi, para in enumerate(shape.text_frame.paragraphs):
+            for ri, run in enumerate(para.runs):
+                if run.text.strip():
+                    runs.append({"slide": si, "shape": shi, "para": pi, "run": ri, "text": run.text})
+with open(r"""\${tmpJson}""", "w", encoding="utf-8") as f:
+    json.dump(runs, f, ensure_ascii=False)
+\`;
+    fs.writeFileSync(tmpExtract, extractScript, "utf-8");
+    const er = spawnSync(py.command, [tmpExtract], { encoding: "utf-8", timeout: 30000 });
+    if (er.status !== 0) throw new Error(er.stderr || "extract failed");
+
+    const runs = JSON.parse(fs.readFileSync(tmpJson, "utf-8"));
+    if (runs.length === 0) {
+      return { success: false, error: "未找到可修改的文本内容（PPT 可能只包含图表或图片）" };
+    }
+
+    // Step 2: LLM editing
+    const textList = runs.map((r, i) => \`[\${i}] Slide\${r.slide + 1}/Shape\${r.shape}/Para\${r.para}/Run\${r.run}: \${r.text}\`).join("\\n");
+    const prompt = \`请按照以下修改指令修改 PowerPoint 文本内容。只修改需要修改的文本 run，不需要修改的保持原文。
+以 JSON 数组格式返回所有 run（包括未修改的），保持原有的 slide/shape/para/run 字段，只更新 text 字段。
+
+修改指令：\${instruction}
+
+当前文本 run：
+\${textList}
+
+直接返回 JSON 数组，不要其他说明。\`;
+    const llmResult = callLLM(prompt);
+    if (!llmResult) {
+      return { success: false, error: "LLM 调用失败", hint: "运行 chainlesschain llm test 检查连接" };
+    }
+
+    let updatedRuns;
+    try {
+      const jsonMatch = llmResult.match(/\\[\\s*\\{[\\s\\S]*\\}\\s*\\]/);
+      updatedRuns = JSON.parse(jsonMatch ? jsonMatch[0] : llmResult);
+    } catch (_e) {
+      return { success: false, error: "LLM 返回格式不正确，无法解析 JSON", hint: "请重试或简化修改指令" };
+    }
+
+    // Step 3: apply changes (only .text, preserve all font/style)
+    const outputFile = buildOutputPath(inputFile, outputDir);
+    const applyScript = \`
+import json, shutil
+from pptx import Presentation
+shutil.copy2(r"""\${inputFile}""", r"""\${outputFile}""")
+prs = Presentation(r"""\${outputFile}""")
+updated = json.loads('''
+\${JSON.stringify(updatedRuns).replace(/'/g, "\\\\'")}
+''')
+for item in updated:
+    slide = prs.slides[item["slide"]]
+    shape = slide.shapes[item["shape"]]
+    if shape.has_text_frame:
+        run = shape.text_frame.paragraphs[item["para"]].runs[item["run"]]
+        run.text = item["text"]
+prs.save(r"""\${outputFile}""")
+\`;
+    fs.writeFileSync(tmpApply, applyScript, "utf-8");
+    const ar = spawnSync(py.command, [tmpApply], { encoding: "utf-8", timeout: 30000 });
+    if (ar.status !== 0) throw new Error(ar.stderr || "apply failed");
+
+    return {
+      success: true,
+      input: inputFile,
+      output: outputFile,
+      action: action || "edit",
+      message: \`PPTX edited (charts/images preserved): \${outputFile}\`,
+      runsModified: updatedRuns.length,
+    };
+  } catch (err) {
+    return { success: false, error: \`PPTX 修改失败: \${err.message}\` };
+  } finally {
+    try { fs.unlinkSync(tmpExtract); } catch (_e) { /* ignore */ }
+    try { fs.unlinkSync(tmpApply); } catch (_e) { /* ignore */ }
+    try { fs.unlinkSync(tmpJson); } catch (_e) { /* ignore */ }
+  }
+}
+
+// ── Main export ───────────────────────────────────────────────────────────────
+async function docEdit(params) {
+  const { input_file, instruction, action = "edit", section, output_dir } = params || {};
+
+  if (!input_file) {
+    return { success: false, error: "Missing required parameter: input_file" };
+  }
+  if (!instruction) {
+    return { success: false, error: "Missing required parameter: instruction" };
+  }
+  if (!fs.existsSync(input_file)) {
+    return { success: false, error: \`Input file not found: \${input_file}\` };
+  }
+
+  const validActions = ["edit", "append", "rewrite-section"];
+  if (!validActions.includes(action)) {
+    return { success: false, error: \`Unsupported action: \${action}\`, hint: \`Supported actions: \${validActions.join(", ")}\` };
+  }
+
+  const ext = path.extname(input_file).toLowerCase();
+
+  if ([".md", ".txt", ".html", ".htm"].includes(ext)) {
+    return editText(input_file, instruction, action, section, output_dir);
+  }
+  if (ext === ".docx") {
+    return editDocx(input_file, instruction, action, section, output_dir);
+  }
+  if (ext === ".xlsx") {
+    return editXlsx(input_file, instruction, action, output_dir);
+  }
+  if (ext === ".pptx") {
+    return editPptx(input_file, instruction, action, output_dir);
+  }
+
+  return {
+    success: false,
+    error: \`不支持的格式: \${ext}\`,
+    hint: "支持的格式: md, txt, html, docx, xlsx, pptx",
+  };
+}
+
+docEdit.execute = async (task, _ctx, _skill) => {
+  const input = typeof task === "string" ? task : (task.input || task.params?.input || "");
+  let p = {};
+  try { p = input.trim().startsWith("{") ? JSON.parse(input) : { input_file: input }; }
+  catch { p = { input_file: input }; }
+  return docEdit(p);
+};
+module.exports = docEdit;
+`,
+  },
+});
+
+const DOC_TEMPLATES_README = `# Document Templates
+
+此目录存放 AI 文档生成的模板文件。
+
+## 使用 AI 生成文档
+
+\`\`\`bash
+# 生成分析报告（Markdown）
+chainlesschain skill run doc-generate "2026年技术趋势分析"
+
+# 生成项目方案（DOCX，需要 pandoc 或 LibreOffice）
+chainlesschain skill run doc-generate "电商平台重构方案" --args '{"style":"proposal","format":"docx"}'
+
+# 生成 PDF（需要 LibreOffice）
+chainlesschain skill run doc-generate "季度运营报告" --args '{"format":"pdf"}'
+
+# 提供大纲精确控制结构
+chainlesschain skill run doc-generate "产品路线图" --args '{"outline":"1.当前状态 2.Q1目标 3.Q2目标 4.风险与缓解 5.资源需求"}'
+\`\`\`
+
+## LibreOffice 格式转换
+
+\`\`\`bash
+# 将 Word 文档转换为 PDF
+chainlesschain skill run libre-convert "report.docx"
+
+# 将 Markdown 转换为 DOCX
+chainlesschain skill run libre-convert "readme.md" --args '{"format":"docx"}'
+\`\`\`
+
+## cli-anything 集成
+
+LibreOffice 具有完整的 CLI 接口（soffice --headless），可以通过 cli-anything 注册以获得完整功能访问：
+
+\`\`\`bash
+# 注册 LibreOffice CLI（适用于高级用途：宏、模板、批量操作）
+chainlesschain cli-anything scan            # 检测 PATH 中的工具
+chainlesschain cli-anything register soffice
+chainlesschain cli-anything list            # 查看已注册工具
+
+# 注册 pandoc（通用文档转换）
+chainlesschain cli-anything register pandoc
+\`\`\`
+
+适合通过 cli-anything 注册的文档工具（有 CLI 接口）：
+- LibreOffice / soffice（全格式转换 + 宏执行）
+- pandoc（多格式转换，尤其 md→docx）
+- wkhtmltopdf（HTML→PDF 高保真）
+
+> **设计边界**：LibreOffice 的日常转换和 AI 文档生成使用 workspace 技能（doc-generate / libre-convert）；
+> 需要直接控制 LibreOffice 高级功能（宏、模板样式、批量脚本）时，通过 cli-anything 注册 soffice 更灵活。
+
+## 自定义模板
+
+在此目录中创建 \`.md\` 模板文件，并在 \`doc-generate\` 技能中通过 outline 参数引用：
+
+\`\`\`bash
+# 创建自定义模板结构
+cat templates/weekly-report-template.md
+# 1.本周完成事项 2.下周计划 3.风险与阻塞 4.需要支持
+
+chainlesschain skill run doc-generate "2026-W12 周报" \\
+  --args '{"outline":"1.本周完成事项 2.下周计划 3.风险与阻塞 4.需要支持","style":"report"}'
+\`\`\`
+`;
+
 const TEMPLATES = {
   "code-project": {
     description:
@@ -150,6 +2250,135 @@ const TEMPLATES = {
       toolsDisabled: [],
     },
   },
+  "ai-doc-creator": {
+    description:
+      "AI 文档创作项目，集成 LibreOffice 格式转换与 AI 驱动的文档生成（报告/方案/说明书）",
+    rules: `# AI 文档创作项目规则
+
+## 文档生成
+
+- 使用 \`chainlesschain skill run doc-generate "主题"\` 让 AI 生成结构化文档
+- 支持输出格式：md（默认）/ html / docx / pdf
+- 文档模板和大纲保存至 \`templates/\` 目录，通过 \`--args '{"outline":"..."}'\` 引用
+
+## LibreOffice 集成
+
+- 格式转换使用 \`chainlesschain skill run libre-convert <file>\`
+- LibreOffice 默认 URL：本地安装（soffice --headless）
+- DOCX 转换优先使用 pandoc（安装：\`winget install pandoc\`）
+- PDF 导出优先使用 LibreOffice（安装：\`winget install LibreOffice.LibreOffice\`）
+
+## cli-anything 集成说明
+
+LibreOffice 具有完整的 CLI 接口（\`soffice --headless\`），**适合通过 cli-anything 注册**以获得高级功能访问：
+
+\`\`\`bash
+# 日常转换 → 使用内置技能
+chainlesschain skill run libre-convert "report.docx"
+
+# 高级操作（宏/模板/批量脚本）→ 通过 cli-anything 注册
+chainlesschain cli-anything register soffice  # 注册完整 LibreOffice CLI
+chainlesschain cli-anything register pandoc   # 注册 pandoc
+\`\`\`
+
+适合 cli-anything 注册的文档工具：
+- \`soffice\` / \`libreoffice\`（LibreOffice，全功能 CLI）
+- \`pandoc\`（通用文档转换）
+- \`wkhtmltopdf\`（HTML→PDF 高保真）
+
+## AI 文档修改
+
+- 使用 \`chainlesschain skill run doc-edit\` 修改现有文档内容
+- xlsx/pptx 修改需要 Python + openpyxl/python-pptx（公式/图表/样式完全保留）
+- 输出命名规则：\`{原文件名}_edited.{扩展名}\`，永不覆盖原文件
+
+## AI 助手行为准则
+
+- 优先使用 doc-generate 技能生成内容，libre-convert 技能转换格式，doc-edit 技能修改内容
+- 批量文档任务推荐使用 agent 模式：\`chainlesschain agent\`
+- 大型文档（>10000字）建议分章节生成，再合并
+- 修改 xlsx/pptx 前确认用户已安装 openpyxl/python-pptx
+- 涉及敏感数据的文档，提示用户注意本地 LLM 模式
+`,
+    skills: ["doc-generate", "libre-convert", "doc-edit", "summarize"],
+    persona: {
+      name: "AI文档助手",
+      role: "你是一个专业的AI文档创作助手，擅长生成各类结构化文档（报告、方案、说明书、README等），熟悉 LibreOffice 文档格式转换和 pandoc 文档处理。你能根据用户需求自动规划文档结构，生成专业、清晰的内容。",
+      behaviors: [
+        "根据用户描述自动选择合适的文档风格（报告/方案/说明书/README）",
+        "主动询问文档目标读者和使用场景以优化内容",
+        "批量任务前确认 LibreOffice 已安装或告知安装方式",
+        "对长文档建议分章节生成以确保质量",
+      ],
+      toolsPriority: ["run_shell", "write_file", "read_file"],
+      toolsDisabled: [],
+    },
+    generateSkills: ["doc-generate", "libre-convert", "doc-edit"],
+    generateDir: "templates",
+    generateDirReadme: "DOC_TEMPLATES_README",
+  },
+  "ai-media-creator": {
+    description:
+      "AI音视频创作项目，集成ComfyUI图像/视频生成与AI音频合成（TTS/音乐）",
+    rules: `# AI 音视频创作项目规则
+
+## ComfyUI 配置
+
+- ComfyUI 服务地址：\`COMFYUI_URL\` 环境变量（默认 http://localhost:8188）
+- 工作流文件保存至项目 \`workflows/\` 目录（API Format JSON 格式）
+- 使用 \`chainlesschain skill run comfyui-image\` 进行文生图
+- 使用 \`chainlesschain skill run comfyui-video\` 进行 AnimateDiff 视频生成
+
+## 音频生成
+
+- TTS 优先使用 \`edge-tts\`（免费）：\`pip install edge-tts\`
+- 高质量 TTS 可配置 \`ELEVENLABS_API_KEY\` 或 \`OPENAI_API_KEY\`
+- 使用 \`chainlesschain skill run audio-gen\` 合成语音
+
+## 工作流规范
+
+- 所有 ComfyUI 工作流以 API Format 保存在 \`workflows/\` 目录
+- 使用有意义的文件名（如 \`txt2img-sd15.json\`、\`animatediff-v3.json\`）
+- 批量创作任务推荐使用 agent 模式：\`chainlesschain agent\`
+
+## cli-anything 集成说明
+
+当用户安装了有 CLI 接口的 AI 工具时，可通过以下命令注册到 ChainlessChain：
+
+\`\`\`bash
+chainlesschain cli-anything register <tool-name>
+\`\`\`
+
+适合注册的工具（有 CLI 接口）：
+- FFmpeg（\`ffmpeg\`）
+- yt-dlp（\`yt-dlp\`）
+- 第三方 ComfyUI CLI 包装脚本
+- audiogen-cli、bark-cli 等音频生成 CLI
+
+注意：ComfyUI 本身以 REST API 为主，不适合通过 cli-anything 注册，
+请直接使用 \`comfyui-image\` 和 \`comfyui-video\` 技能。
+
+## AI 助手行为准则
+
+- 优先使用项目 workflows/ 中已保存的工作流
+- 批量任务使用 agent 模式以获得最佳控制
+- 生成大文件时提示用户确认存储空间
+`,
+    skills: ["comfyui-image", "comfyui-video", "audio-gen", "summarize"],
+    persona: {
+      name: "AI创作助手",
+      role: "你是一个专业的AI音视频创作助手，帮助用户利用 ComfyUI、AnimateDiff 和 AI 音频合成工具创作高质量的图像、视频和音频内容。你熟悉 Stable Diffusion 生图技术、提示词工程和 AnimateDiff 动画生成。",
+      behaviors: [
+        "根据用户创作需求推荐合适的工作流和参数",
+        "提供专业的 Stable Diffusion 提示词建议",
+        "在批量任务前确认存储空间和 ComfyUI 连接状态",
+        "推荐免费开源工具（edge-tts、piper-tts）优先于付费 API",
+      ],
+      toolsPriority: ["run_shell", "write_file", "read_file"],
+      toolsDisabled: [],
+    },
+    generateSkills: ["comfyui-image", "comfyui-video", "audio-gen"],
+  },
   empty: {
     description: "Bare project with minimal configuration",
     rules: `# Project Rules
@@ -169,7 +2398,7 @@ export function registerInitCommand(program) {
     )
     .option(
       "-t, --template <name>",
-      "Project template (code-project, data-science, devops, medical-triage, agriculture-expert, general-assistant, empty)",
+      "Project template (code-project, data-science, devops, medical-triage, agriculture-expert, general-assistant, ai-media-creator, ai-doc-creator, empty)",
       "empty",
     )
     .option("-y, --yes", "Skip prompts, use defaults")
@@ -276,6 +2505,41 @@ ${tmpl.persona.behaviors?.map((b) => `- ${b}`).join("\n") || ""}
           );
         }
 
+        // Generate additional workspace skills for templates that require them
+        if (tmpl.generateSkills) {
+          for (const skillName of tmpl.generateSkills) {
+            const skillTmpl = SKILL_TEMPLATES[skillName];
+            if (!skillTmpl) continue;
+            const skillDir = path.join(ccDir, "skills", skillName);
+            fs.mkdirSync(skillDir, { recursive: true });
+            fs.writeFileSync(
+              path.join(skillDir, "SKILL.md"),
+              skillTmpl.md,
+              "utf-8",
+            );
+            fs.writeFileSync(
+              path.join(skillDir, "handler.js"),
+              skillTmpl.handler,
+              "utf-8",
+            );
+          }
+          // Create companion directory (workflows/ or templates/)
+          const dirName = tmpl.generateDir || "workflows";
+          const readmeContent =
+            tmpl.generateDirReadme === "DOC_TEMPLATES_README"
+              ? DOC_TEMPLATES_README
+              : WORKFLOW_README;
+          const companionDir = path.join(cwd, dirName);
+          if (!fs.existsSync(companionDir)) {
+            fs.mkdirSync(companionDir, { recursive: true });
+            fs.writeFileSync(
+              path.join(companionDir, "README.md"),
+              readmeContent,
+              "utf-8",
+            );
+          }
+        }
+
         logger.success(
           `Initialized ChainlessChain project in ${chalk.cyan(cwd)}`,
         );
@@ -284,6 +2548,15 @@ ${tmpl.persona.behaviors?.map((b) => `- ${b}`).join("\n") || ""}
         logger.log(`  Config:    ${chalk.gray(".chainlesschain/config.json")}`);
         logger.log(`  Rules:     ${chalk.gray(".chainlesschain/rules.md")}`);
         logger.log(`  Skills:    ${chalk.gray(".chainlesschain/skills/")}`);
+        if (tmpl.generateSkills) {
+          const dirLabel = tmpl.generateDir || "workflows";
+          logger.log(
+            `  Skills:    ${chalk.gray(tmpl.generateSkills.map((s) => `.chainlesschain/skills/${s}/`).join(", "))}`,
+          );
+          logger.log(
+            `  ${dirLabel.charAt(0).toUpperCase() + dirLabel.slice(1)}: ${chalk.gray(`${dirLabel}/`)}`,
+          );
+        }
         logger.log("");
         logger.log(chalk.bold("Next steps:"));
         logger.log(
@@ -295,6 +2568,38 @@ ${tmpl.persona.behaviors?.map((b) => `- ${b}`).join("\n") || ""}
         logger.log(
           `  ${chalk.cyan("chainlesschain agent")}             Start the AI agent`,
         );
+        if (tmpl.generateSkills) {
+          logger.log("");
+          if (template === "ai-media-creator") {
+            logger.log(chalk.bold("AI Media Setup:"));
+            logger.log(
+              `  ${chalk.cyan("pip install edge-tts")}                Install free TTS backend`,
+            );
+            logger.log(
+              `  ${chalk.cyan("chainlesschain skill run comfyui-image")} \\"a sunset\\"  Generate image`,
+            );
+            logger.log(
+              `  ${chalk.cyan("chainlesschain skill run audio-gen")} \\"Hello world\\"  Synthesize speech`,
+            );
+          } else if (template === "ai-doc-creator") {
+            logger.log(chalk.bold("AI Doc Setup:"));
+            logger.log(
+              `  ${chalk.cyan("chainlesschain skill run doc-generate")} \\"My Report\\"  Generate AI document`,
+            );
+            logger.log(
+              `  ${chalk.cyan("chainlesschain skill run libre-convert")} \\"file.docx\\"  Convert to PDF`,
+            );
+            logger.log(
+              `  ${chalk.cyan("chainlesschain skill run doc-edit")} \\"file.md\\"    Edit existing document with AI`,
+            );
+            logger.log(
+              `  ${chalk.cyan("winget install pandoc")}              Install pandoc for DOCX output`,
+            );
+          }
+          logger.log(
+            `  ${chalk.cyan("chainlesschain cli-anything scan")}    Scan for CLI tools to register`,
+          );
+        }
         logger.log("");
       } catch (err) {
         logger.error(`Failed to initialize: ${err.message}`);