Merge pull request #6 from echoVic/feature/skills-enhancement

Feature/skills enhancement

Author: echoVic

.blade/settings.local.json

@@ -5,7 +5,26 @@
       "Bash(git status)",
       "Bash(git diff *)",
       "Bash(ls -l)",
-      "Write(**/*.md)"
+      "Write(**/*.md)",
+      "Bash(ls -la *)",
+      "Bash(bun install *)",
+      "Skill",
+      "Write(**/*.py)",
+      "Bash(python test_echovic_website.py)",
+      "Bash(pip install *)",
+      "Bash(playwright install *)",
+      "Bash(python test_echovic_comprehensive.py)",
+      "Bash(curl -s *)",
+      "Bash(curl -I *)",
+      "Edit(**/*.py)",
+      "Bash(python -c *)",
+      "Bash(python test_echovic_debug.py)",
+      "Bash(python test_echovic_fixed.py)",
+      "Bash(python test_echovic_wait.py)",
+      "Bash(python test_echovic_basic.py)",
+      "Bash(python test_echovic_individual.py)",
+      "Bash(python test_echovic_analysis.py)",
+      "Bash(python3 -c *)"
     ],
     "ask": [],
     "deny": []

.claude/settings.local.json

@@ -71,7 +71,11 @@
       "WebFetch(domain:www.npmjs.com)",
       "WebFetch(domain:deepwiki.com)",
       "Bash(timeout 3 npm run start:*)",
-      "Bash(gtimeout 3 npm run start:*)"
+      "Bash(gtimeout 3 npm run start:*)",
+      "WebFetch(domain:leehanchung.github.io)",
+      "WebFetch(domain:simonwillison.net)",
+      "WebFetch(domain:scottspence.com)",
+      "WebFetch(domain:docs.anthropic.com)"
     ],
     "deny": [],
     "ask": [],

src/agent/Agent.ts

@@ -58,10 +58,21 @@ import type {
   LoopResult,
   UserMessageContent,
 } from './types.js';
+import { discoverSkills, injectSkillsMetadata } from '../skills/index.js';
 
 // 创建 Agent 专用 Logger
 const logger = createLogger(LogCategory.AGENT);
 
+/**
+ * Skill 执行上下文
+ * 用于跟踪当前活动的 Skill 及其工具限制
+ */
+interface SkillExecutionContext {
+  skillName: string;
+  allowedTools?: string[];
+  basePath: string;
+}
+
 export class Agent {
   private config: BladeConfig;
   private runtimeOptions: AgentOptions;
@@ -76,6 +87,9 @@ export class Agent {
   private executionEngine!: ExecutionEngine;
   private attachmentCollector?: AttachmentCollector;
 
+  // Skill 执行上下文（用于 allowed-tools 限制）
+  private activeSkillContext?: SkillExecutionContext;
+
   constructor(
     config: BladeConfig,
     runtimeOptions: AgentOptions = {},
@@ -171,7 +185,10 @@ export class Agent {
       // 3. 加载 subagent 配置
       await this.loadSubagents();
 
-      // 4. 初始化核心组件
+      // 4. 发现并注册 Skills
+      await this.discoverSkills();
+
+      // 5. 初始化核心组件
       // 获取当前模型配置（从 Store）
       const modelConfig = getCurrentModel();
       if (!modelConfig) {
@@ -468,7 +485,11 @@ IMPORTANT: Execute according to the approved plan above. Follow the steps exactl
       // 根据 permissionMode 决定工具暴露策略（单一信息源：ToolRegistry.getFunctionDeclarationsByMode）
       const registry = this.executionPipeline.getRegistry();
       const permissionMode = context.permissionMode as PermissionMode | undefined;
-      const tools = registry.getFunctionDeclarationsByMode(permissionMode);
+      let rawTools = registry.getFunctionDeclarationsByMode(permissionMode);
+      // 注入 Skills 元数据到 Skill 工具的 <available_skills> 占位符
+      rawTools = injectSkillsMetadata(rawTools);
+      // 应用 Skill 的 allowed-tools 限制（如果有活动的 Skill）
+      const tools = this.applySkillToolRestrictions(rawTools);
       const isPlanMode = permissionMode === PermissionMode.PLAN;
 
       if (isPlanMode) {
@@ -1028,6 +1049,24 @@ IMPORTANT: Execute according to the approved plan above. Follow the steps exactl
               );
             }
 
+            // 如果是 Skill 工具，设置执行上下文（用于 allowed-tools 限制）
+            if (toolCall.function.name === 'Skill' && result.success && result.metadata) {
+              const metadata = result.metadata as Record<string, unknown>;
+              if (metadata.skillName) {
+                this.activeSkillContext = {
+                  skillName: metadata.skillName as string,
+                  allowedTools: metadata.allowedTools as string[] | undefined,
+                  basePath: (metadata.basePath as string) || '',
+                };
+                logger.debug(
+                  `🎯 Skill "${this.activeSkillContext.skillName}" activated` +
+                    (this.activeSkillContext.allowedTools
+                      ? ` with allowed tools: ${this.activeSkillContext.allowedTools.join(', ')}`
+                      : '')
+                );
+              }
+            }
+
             // 添加工具执行结果到消息历史
             // 优先使用 llmContent（为 LLM 准备的详细内容），displayContent 仅用于终端显示
             let toolResultContent = result.success
@@ -1673,6 +1712,92 @@ IMPORTANT: Execute according to the approved plan above. Follow the steps exactl
     }
   }
 
+  /**
+   * 发现并注册 Skills
+   * Skills 是动态 Prompt 扩展机制，允许 AI 根据用户请求自动调用专业能力
+   */
+  private async discoverSkills(): Promise<void> {
+    try {
+      const result = await discoverSkills({
+        cwd: process.cwd(),
+      });
+
+      if (result.skills.length > 0) {
+        logger.debug(
+          `✅ Discovered ${result.skills.length} skills: ${result.skills.map((s) => s.name).join(', ')}`
+        );
+      } else {
+        logger.debug('📦 No skills configured');
+      }
+
+      // 记录发现过程中的错误（不阻塞初始化）
+      for (const error of result.errors) {
+        logger.warn(`⚠️  Skill loading error at ${error.path}: ${error.error}`);
+      }
+    } catch (error) {
+      logger.warn('Failed to discover skills:', error);
+      // 不抛出错误，允许 Agent 继续初始化
+    }
+  }
+
+  /**
+   * 应用 Skill 的 allowed-tools 限制
+   * 如果有活动的 Skill 且定义了 allowed-tools，则过滤可用工具列表
+   *
+   * @param tools - 原始工具列表
+   * @returns 过滤后的工具列表
+   */
+  private applySkillToolRestrictions(
+    tools: import('../tools/types/index.js').FunctionDeclaration[]
+  ): import('../tools/types/index.js').FunctionDeclaration[] {
+    // 如果没有活动的 Skill，或者 Skill 没有定义 allowed-tools，返回原始工具列表
+    if (!this.activeSkillContext?.allowedTools) {
+      return tools;
+    }
+
+    const allowedTools = this.activeSkillContext.allowedTools;
+    logger.debug(
+      `🔒 Applying Skill tool restrictions: ${allowedTools.join(', ')}`
+    );
+
+    // 过滤工具列表，只保留 allowed-tools 中指定的工具
+    const filteredTools = tools.filter((tool) => {
+      // 检查工具名称是否在 allowed-tools 列表中
+      // 支持精确匹配和通配符模式（如 Bash(git:*)）
+      return allowedTools.some((allowed) => {
+        // 精确匹配
+        if (allowed === tool.name) {
+          return true;
+        }
+
+        // 通配符匹配：Bash(git:*) 匹配 Bash
+        const match = allowed.match(/^(\w+)\(.*\)$/);
+        if (match && match[1] === tool.name) {
+          return true;
+        }
+
+        return false;
+      });
+    });
+
+    logger.debug(
+      `🔒 Filtered tools: ${filteredTools.map((t) => t.name).join(', ')} (${filteredTools.length}/${tools.length})`
+    );
+
+    return filteredTools;
+  }
+
+  /**
+   * 清除 Skill 执行上下文
+   * 当 Skill 执行完成或需要重置时调用
+   */
+  public clearSkillContext(): void {
+    if (this.activeSkillContext) {
+      logger.debug(`🎯 Skill "${this.activeSkillContext.skillName}" deactivated`);
+      this.activeSkillContext = undefined;
+    }
+  }
+
   /**
    * 处理 @ 文件提及（支持纯文本和多模态消息）
    * 从用户消息中提取 @ 提及，读取文件内容，并追加到消息
@@ -1781,7 +1906,7 @@ IMPORTANT: Execute according to the approved plan above. Follow the steps exactl
   }
 
   /**
-   * 处理 @ 文件提及（纯文本版本）
+   * 处理 @ 文件提及
    * 从用户消息中提取 @ 提及，读取文件内容，并追加到消息
    *
    * @param message - 原始用户消息

src/prompts/builder.ts

@@ -16,9 +16,13 @@
 import { promises as fs } from 'fs';
 import path from 'path';
 import { PermissionMode } from '../config/types.js';
+import { getSkillRegistry } from '../skills/index.js';
 import { getEnvironmentContext } from '../utils/environment.js';
 import { DEFAULT_SYSTEM_PROMPT, PLAN_MODE_SYSTEM_PROMPT } from './default.js';
 
+/** available_skills 占位符的正则表达式 */
+const AVAILABLE_SKILLS_REGEX = /<available_skills>\s*<\/available_skills>/;
+
 /**
  * 提示词构建选项
  */
@@ -141,11 +145,33 @@ export async function buildSystemPrompt(
   }
 
   // 组合各部分
-  const prompt = parts.join('\n\n---\n\n');
+  let prompt = parts.join('\n\n---\n\n');
+
+  // 注入 Skills 元数据到 <available_skills> 占位符
+  prompt = injectSkillsToPrompt(prompt);
 
   return { prompt, sources };
 }
 
+/**
+ * 注入 Skills 列表到系统提示的 <available_skills> 占位符
+ */
+function injectSkillsToPrompt(prompt: string): string {
+  const registry = getSkillRegistry();
+  const skillsList = registry.generateAvailableSkillsList();
+
+  // 如果没有 skills，保持占位符为空（但保留标签结构）
+  if (!skillsList) {
+    return prompt;
+  }
+
+  // 替换占位符
+  return prompt.replace(
+    AVAILABLE_SKILLS_REGEX,
+    `<available_skills>\n${skillsList}\n</available_skills>`
+  );
+}
+
 /**
  * 加载项目 BLADE.md 配置
  */

src/prompts/default.ts

@@ -140,6 +140,16 @@ The user will primarily request you perform software engineering tasks. This inc
 - You can call multiple tools in a single response. Make independent tool calls in parallel. If calls depend on previous results, run them sequentially. Never use placeholders or guess missing parameters.
 - Use specialized tools instead of bash commands: Read for files, Edit for editing, Write for creating. Reserve Bash for system commands only.
 
+## Skills
+When users ask you to perform tasks, check if any of the available skills below can help complete the task more effectively. Skills provide specialized capabilities and domain knowledge.
+
+How to invoke skills:
+- Use the Skill tool with the skill name
+- Example: \`skill: "commit-message"\` to generate commit messages
+
+<available_skills>
+</available_skills>
+
 ## Code References
 
 When referencing specific functions or pieces of code include the pattern \`file_path:line_number\` to allow the user to easily navigate to the source code location.