feat: integrate skills support

README.md

@@ -13,8 +13,9 @@
 - 🚀 **Code-as-Action**: Agent generates and executes code to complete tasks, rather than just calling predefined tools
 - 🔒 **Secure Sandbox**: AI-generated code runs safely in GraalVM polyglot sandbox with resource isolation
 - 📊 **Multi-dimensional Evaluation**: Multi-layer intent recognition through Evaluation Graph, precisely guiding Agent behavior
-- 🔄 **Dynamic Prompt Builder**: Dynamically inject context (experiences, knowledge, etc.) into prompts based on scenarios and evaluation results
-- 🧠 **Experience Learning**: Automatically accumulates successful experiences to continuously improve performance on subsequent tasks
+- 🔄 **Dynamic Prompt Builder**: Dynamically inject runtime context, prefetched experience candidates, and stable guidance into prompts based on scenarios and evaluation results
+- 🧠 **Unified Experience System**: Manage COMMON / REACT / TOOL experiences in one model, support conversion to and from the Skills model, and improve experience efficiency and quality through progressive disclosure
+- 🗂️ **Management Console**: Provide a dedicated management module for experience search, CRUD, statistics, and SKILL preview / import / export, so reusable experience and skill assets can be maintained in one place
 - ⚡ **Fast Response**: For familiar scenarios, skip LLM reasoning process and respond quickly based on experience
 
 ## 📖 Introduction
@@ -29,9 +30,10 @@
 Assistant Agent is a fully-featured intelligent assistant with the following core capabilities:
 
 - 🔍 **Intelligent Q&A**: Supports unified retrieval architecture across multiple data sources (extensible via SPI for knowledge base, Web, etc.), providing accurate, traceable answers
-- 🛠️ **Tool Invocation**: Supports MCP, HTTP API (OpenAPI) and other protocols, flexibly access massive tools, combine multiple tools to implement complex business workflows
+- 🛠️ **Tool Invocation**: Supports MCP, HTTP API (OpenAPI) and other protocols, can invoke tools directly in the React loop or orchestrate multiple tools in generated code for more complex workflows
 - ⏰ **Proactive Service**: Supports scheduled tasks, delayed execution, event callbacks, letting the assistant proactively serve you
 - 📬 **Multi-channel Delivery**: Built-in IDE reply, extensible to DingTalk, Feishu, WeCom, Webhook and other channels via SPI
+- 🧩 **Experience and Operations Management**: Supports experience management, tenant-aware retrieval, and bidirectional conversion between the experience model and SKILL packages, making it easier to accumulate and reuse business capabilities
 
 ### Why Choose Assistant Agent?
 
@@ -69,14 +71,15 @@ AssistantAgent/
 ├── assistant-agent-core            # Core engine: GraalVM executor, tool registry
 ├── assistant-agent-extensions      # Extension modules:
 │   ├── dynamic/               #   - Dynamic tools (MCP, HTTP API)
-│   ├── experience/            #   - Experience management and FastIntent configuration
+│   ├── experience/            #   - Unified experience runtime, disclosure, and FastIntent configuration
 │   ├── learning/              #   - Learning extraction and storage
 │   ├── search/                #   - Unified search capability
 │   ├── reply/                 #   - Multi-channel reply
 │   ├── trigger/               #   - Trigger mechanism
 │   └── evaluation/            #   - Evaluation integration
 ├── assistant-agent-prompt-builder  # Prompt dynamic assembly
 ├── assistant-agent-evaluation      # Evaluation engine
+├── assistant-agent-management      # Experience management and SKILL conversion APIs
 ├── assistant-agent-autoconfigure   # Spring Boot auto-configuration
 └── assistant-agent-start           # Startup module
 ```
@@ -228,8 +231,8 @@ For detailed documentation on each module, please visit our [Documentation Site]
 
 | Module | Description | Documentation |
 |--------|-------------|---------------|
-| **Experience** | Accumulate and reuse historical successful execution experiences with FastIntent support | [Quick Start](https://java2ai.com/agents/assistantagent/features/experience/quickstart) ｜ [Advanced](https://java2ai.com/agents/assistantagent/features/experience/advanced) |
-| **Learning** | Automatically extract valuable experiences from Agent execution history | [Quick Start](https://java2ai.com/agents/assistantagent/features/learning/quickstart) ｜ [Advanced](https://java2ai.com/agents/assistantagent/features/learning/advanced) |
+| **Experience** | Unified COMMON / REACT / TOOL experience model with FastIntent support, conversion to and from Skills, progressive disclosure, and runtime retrieval via `search_exp` / `read_exp` | [Quick Start](https://java2ai.com/agents/assistantagent/features/experience/quickstart) ｜ [Advanced](https://java2ai.com/agents/assistantagent/features/experience/advanced) |
+| **Learning** | Automatically extract valuable COMMON / REACT / TOOL experiences from Agent execution history | [Quick Start](https://java2ai.com/agents/assistantagent/features/learning/quickstart) ｜ [Advanced](https://java2ai.com/agents/assistantagent/features/learning/advanced) |
 | **Search** | Multi-source unified search engine for knowledge-based Q&A | [Quick Start](https://java2ai.com/agents/assistantagent/features/search/quickstart) ｜ [Advanced](https://java2ai.com/agents/assistantagent/features/search/advanced) |
 
 ### Interaction Capabilities
@@ -239,6 +242,13 @@ For detailed documentation on each module, please visit our [Documentation Site]
 | **Reply Channel** | Multi-channel message reply with routing support | [Quick Start](https://java2ai.com/agents/assistantagent/features/reply/quickstart) ｜ [Advanced](https://java2ai.com/agents/assistantagent/features/reply/advanced) |
 | **Trigger** | Scheduled tasks, delayed execution, and event callback triggers | [Quick Start](https://java2ai.com/agents/assistantagent/features/trigger/quickstart) ｜ [Advanced](https://java2ai.com/agents/assistantagent/features/trigger/advanced) |
 
+### Management Capabilities
+
+| Capability | Description | Entry |
+|--------|-------------|-------|
+| **Experience Management API** | Tenant-aware listing, search, stats, and CRUD for runtime experiences | [ExperienceManagementController](assistant-agent-management/src/main/java/com/alibaba/assistant/agent/management/controller/ExperienceManagementController.java) |
+| **SKILL Conversion API** | Convert to and from SKILL content, enabling bidirectional conversion between the Skills model and the unified experience model | [SkillExchangeController](assistant-agent-management/src/main/java/com/alibaba/assistant/agent/management/controller/SkillExchangeController.java) |
+
 ### Additional Resources
 
 | Resource | Link |

README_zh.md

@@ -13,8 +13,9 @@
 - 🚀 **代码即行动（Code-as-Action）**：Agent 通过生成并执行代码来完成任务，而非仅仅调用预定义工具，可以在代码中灵活组合多个工具，实现复杂流程
 - 🔒 **安全沙箱**：AI 生成的代码在 GraalVM 多语言沙箱中安全运行，具备资源隔离能力
 - 📊 **多维评估**：通过评估图（Graph）进行多层次意图识别，精准指导 Agent 行为
-- 🔄 **Prompt 动态组装**：根据场景及前置评估结果动态注入上下文（经验、知识等）到 Prompt 中，灵活处理不同任务
-- 🧠 **经验学习**：自动积累成功经验，持续提升后续任务的表现
+- 🔄 **Prompt 动态组装**：根据场景及前置评估结果，动态注入运行时上下文、预取经验候选和稳定指导信息到 Prompt 中，灵活处理不同任务
+- 🧠 **统一经验体系**：以统一模型管理 COMMON / REACT / TOOL 三类经验，支持与 Skills 模型互相转化，并通过渐进式披露思想提高经验使用效率与效果
+- 🗂️ **管理后台**：通过独立管理模块提供经验检索、CRUD、统计，以及 SKILL 预览 / 导入 / 导出等能力，便于统一维护可复用经验与技能资产
 - ⚡ **快速响应**：熟悉场景下，跳过 LLM 推理过程，基于经验快速响应
 
 ## 📖 简介
@@ -26,9 +27,10 @@
 Assistant Agent 是一个功能完整的智能助手，具备以下核心能力：
 
 - 🔍 **智能问答**：支持多数据源统一检索架构（通过 SPI 可扩展知识库、Web 等数据源），提供准确、可溯源的答案
-- 🛠️ **工具调用**：支持 MCP、HTTP API（OpenAPI）等协议，灵活接入海量工具，可组合调用实现复杂业务流程
+- 🛠️ **工具调用**：支持 MCP、HTTP API（OpenAPI）等协议，既可在 React 阶段直接调用工具，也可在生成代码中组合多个工具完成复杂业务流程
 - ⏰ **主动服务**：支持定时任务、延迟执行、事件回调，让助手主动为你服务
 - 📬 **多渠道触达**：内置 IDE 回复，通过 SPI 可扩展钉钉、飞书、企微、Webhook 等渠道
+- 🧩 **运维与经验管理**：支持经验管理、租户维度检索，以及经验模型与 SKILL 包之间的双向转换，方便沉淀并复用业务能力
 
 ### 为什么选择 Assistant Agent？
 
@@ -64,14 +66,15 @@ AssistantAgent/
 ├── assistant-agent-core            # 核心引擎：GraalVM 执行器、工具注册表
 ├── assistant-agent-extensions      # 扩展模块：
 │   ├── dynamic/               #   - 动态工具（MCP、HTTP API）
-│   ├── experience/            #   - 经验管理与快速意图配置
+│   ├── experience/            #   - 统一经验运行时、经验披露与快速意图配置
 │   ├── learning/              #   - 学习提取与存储
 │   ├── search/                #   - 统一搜索能力
 │   ├── reply/                 #   - 多渠道回复
 │   ├── trigger/               #   - 触发器机制
 │   └── evaluation/            #   - 评估集成
 ├── assistant-agent-prompt-builder  # Prompt 动态组装
 ├── assistant-agent-evaluation      # 评估引擎
+├── assistant-agent-management      # 经验管理与 SKILL 转换 API
 ├── assistant-agent-autoconfigure   # Spring Boot 自动配置
 └── assistant-agent-start           # 启动模块
 ```
@@ -223,8 +226,8 @@ public class MyKnowledgeSearchProvider implements SearchProvider {
 
 | 模块 | 说明 | 文档 |
 |------|------|------|
-| **经验模块** | 积累和复用历史成功执行经验，支持快速意图响应 | [快速开始](https://java2ai.com/agents/assistantagent/features/experience/quickstart) ｜ [高级特性](https://java2ai.com/agents/assistantagent/features/experience/advanced) |
-| **学习模块** | 从 Agent 执行历史中自动提取有价值的经验 | [快速开始](https://java2ai.com/agents/assistantagent/features/learning/quickstart) ｜ [高级特性](https://java2ai.com/agents/assistantagent/features/learning/advanced) |
+| **经验模块** | 基于统一 COMMON / REACT / TOOL 模型管理经验，支持快速意图、与 Skills 模型互转、渐进式披露，以及通过 `search_exp` / `read_exp` 进行运行时检索 | [快速开始](https://java2ai.com/agents/assistantagent/features/experience/quickstart) ｜ [高级特性](https://java2ai.com/agents/assistantagent/features/experience/advanced) |
+| **学习模块** | 从 Agent 执行历史中自动提取有价值的 COMMON / REACT / TOOL 经验 | [快速开始](https://java2ai.com/agents/assistantagent/features/learning/quickstart) ｜ [高级特性](https://java2ai.com/agents/assistantagent/features/learning/advanced) |
 | **搜索模块** | 多数据源统一检索引擎，支持知识问答 | [快速开始](https://java2ai.com/agents/assistantagent/features/search/quickstart) ｜ [高级特性](https://java2ai.com/agents/assistantagent/features/search/advanced) |
 
 ### 交互能力
@@ -234,6 +237,13 @@ public class MyKnowledgeSearchProvider implements SearchProvider {
 | **回复渠道** | 多渠道消息回复，支持渠道路由 | [快速开始](https://java2ai.com/agents/assistantagent/features/reply/quickstart) ｜ [高级特性](https://java2ai.com/agents/assistantagent/features/reply/advanced) |
 | **触发器** | 定时任务、延迟执行、事件回调触发 | [快速开始](https://java2ai.com/agents/assistantagent/features/trigger/quickstart) ｜ [高级特性](https://java2ai.com/agents/assistantagent/features/trigger/advanced) |
 
+### 管理能力
+
+| 能力               | 说明                                         | 入口 |
+|------------------|--------------------------------------------|------|
+| **经验管理 API**     | 提供面向租户的经验列表、搜索、统计与 CRUD 能力                 | [ExperienceManagementController](assistant-agent-management/src/main/java/com/alibaba/assistant/agent/management/controller/ExperienceManagementController.java) |
+| **SKILL 转换 API** | 提供与 SKILL 的转换能力，支撑 Skills 模型与统一经验模型之间的双向转换 | [SkillExchangeController](assistant-agent-management/src/main/java/com/alibaba/assistant/agent/management/controller/SkillExchangeController.java) |
+
 ### 更多资源
 
 | 资源 | 链接 |

assistant-agent-autoconfigure/pom.xml

@@ -6,7 +6,7 @@
     <parent>
         <groupId>com.alibaba.agent.assistant</groupId>
         <artifactId>assistant-agent</artifactId>
-        <version>0.2.4</version>
+        <version>0.2.5</version>
     </parent>
 
     <artifactId>assistant-agent-autoconfigure</artifactId>

assistant-agent-autoconfigure/src/main/java/com/alibaba/assistant/agent/autoconfigure/CodeactAgent.java

@@ -29,7 +29,6 @@
 import com.alibaba.assistant.agent.core.tool.ToolRegistryBridgeFactory;
 import com.alibaba.assistant.agent.core.tool.schema.ReturnSchemaRegistry;
 import com.alibaba.assistant.agent.extension.experience.config.ExperienceExtensionProperties;
-import com.alibaba.assistant.agent.extension.experience.fastintent.CodeFastIntentSupport;
 import com.alibaba.assistant.agent.extension.experience.fastintent.FastIntentService;
 import com.alibaba.assistant.agent.extension.experience.spi.ExperienceProvider;
 import com.alibaba.assistant.agent.autoconfigure.tools.ExecuteCodeTool;
@@ -627,11 +626,6 @@ public CodeactAgent build() {
 				this.executionTimeoutMs
 			);
 
-			CodeFastIntentSupport codeFastIntentSupport = 
-				(experienceProvider != null && experienceExtensionProperties != null && fastIntentService != null)
-					? new CodeFastIntentSupport(experienceProvider, experienceExtensionProperties, fastIntentService)
-					: null;
-
 			ExecuteCodeTool executeCodeTool = new ExecuteCodeTool(this.executor, this.codeContext, this.variableProvider);
 
 			// Manually create the components like DefaultBuilder does
@@ -706,10 +700,10 @@ public CodeactAgent build() {
 			}
 
 			allTools.add(WriteCodeTool.createWriteCodeToolCallback(
-				this.codeContext, this.environmentManager, codeFastIntentSupport));
+				this.codeContext, this.environmentManager));
 
 			allTools.add(WriteConditionCodeTool.createWriteConditionCodeToolCallback(
-				this.codeContext, this.environmentManager, codeFastIntentSupport));
+				this.codeContext, this.environmentManager));
 
 			allTools.add(
 				FunctionToolCallback.builder("execute_code", executeCodeTool)

assistant-agent-autoconfigure/src/main/java/com/alibaba/assistant/agent/autoconfigure/evaluation/DefaultEvaluationSuiteConfig.java

@@ -18,15 +18,13 @@
 import com.alibaba.assistant.agent.evaluation.DefaultEvaluationService;
 import com.alibaba.assistant.agent.evaluation.EvaluationService;
 import com.alibaba.assistant.agent.evaluation.builder.EvaluationSuiteBuilder;
+import com.alibaba.assistant.agent.evaluation.evaluator.Evaluator;
 import com.alibaba.assistant.agent.evaluation.evaluator.EvaluatorRegistry;
 import com.alibaba.assistant.agent.evaluation.evaluator.LLMBasedEvaluator;
-import com.alibaba.assistant.agent.evaluation.evaluator.RuleBasedEvaluator;
 import com.alibaba.assistant.agent.evaluation.model.EvaluationCriterion;
 import com.alibaba.assistant.agent.evaluation.model.EvaluationSuite;
 import com.alibaba.assistant.agent.extension.evaluation.config.CodeactEvaluationContextFactory;
-import com.alibaba.assistant.agent.extension.evaluation.experience.ExperienceRetrievalEvaluatorFactory;
 import com.alibaba.assistant.agent.extension.evaluation.hook.ReactBeforeModelEvaluationHook;
-import com.alibaba.assistant.agent.extension.experience.spi.ExperienceProvider;
 import com.alibaba.assistant.agent.extension.prompt.ReactPromptContributorModelHook;
 import com.alibaba.assistant.agent.prompt.PromptContributorManager;
 import com.alibaba.cloud.ai.graph.agent.hook.Hook;
@@ -74,18 +72,18 @@ public class DefaultEvaluationSuiteConfig {
     private final DefaultEvaluationProperties properties;
     private final ChatModel chatModel;
     private final List<EvaluationCriterionProvider> criterionProviders;
-    private final ExperienceProvider experienceProvider;
+    private final List<Evaluator> customEvaluators;
 
     public DefaultEvaluationSuiteConfig(
             DefaultEvaluationProperties properties,
             ChatModel chatModel,
             @Autowired(required = false) List<EvaluationCriterionProvider> criterionProviders,
-            @Autowired(required = false) ExperienceProvider experienceProvider) {
+            @Autowired(required = false) List<Evaluator> customEvaluators) {
         this.properties = properties;
         this.chatModel = chatModel;
         this.criterionProviders = criterionProviders;
-        this.experienceProvider = experienceProvider;
-        log.info("DefaultEvaluationSuiteConfig#<init> - reason=初始化默认评估套件配置, experienceProviderAvailable={}", experienceProvider != null);
+        this.customEvaluators = customEvaluators;
+        log.info("DefaultEvaluationSuiteConfig#<init> - reason=初始化默认评估套件配置");
     }
 
     /**
@@ -178,9 +176,7 @@ private EvaluationSuite createDefaultSuite() {
     /**
      * 创建默认的评估器注册表
      *
-     * <p>Starter 层自动装配：
-     * - LLM 评估器
-     * - 经验检索评估器（如果 ExperienceProvider 可用）
+     * <p>Starter 层自动装配 LLM 评估器。
      */
     private EvaluatorRegistry createDefaultEvaluatorRegistry() {
         EvaluatorRegistry registry = new EvaluatorRegistry();
@@ -189,14 +185,15 @@ private EvaluatorRegistry createDefaultEvaluatorRegistry() {
         LLMBasedEvaluator llmEvaluator = new LLMBasedEvaluator(chatModel, "llm-based");
         registry.registerEvaluator(llmEvaluator);
 
-
-        // 自动注册经验检索评估器（如果 ExperienceProvider 可用）
-        if (experienceProvider != null && properties.getExperience().isEnabled()) {
-            int maxExperiences = properties.getExperience().getMaxExperiencesPerType();
-            RuleBasedEvaluator expEvaluator = ExperienceRetrievalEvaluatorFactory.createExperienceEvaluator(
-                    experienceProvider, maxExperiences);
-            registry.registerEvaluator(expEvaluator);
-            log.info("DefaultEvaluationSuiteConfig#createDefaultEvaluatorRegistry - reason=注册经验检索评估器");
+        if (customEvaluators != null) {
+            for (Evaluator evaluator : customEvaluators) {
+                if (evaluator == null || "llm-based".equals(evaluator.getEvaluatorId())) {
+                    continue;
+                }
+                registry.registerEvaluator(evaluator);
+                log.info("DefaultEvaluationSuiteConfig#createDefaultEvaluatorRegistry - reason=注册自定义评估器, evaluatorId={}",
+                        evaluator.getEvaluatorId());
+            }
         }
 
         return registry;
@@ -216,4 +213,3 @@ public EvaluatorRegistry evaluatorRegistry() {
         return createDefaultEvaluatorRegistry();
     }
 }
-