--- name: discover-agent description: 根据用户需求推荐最匹配的智能体，展示候选列表并引导选择。Use when 用户描述需求但不确定该找哪个智能体帮忙，或想浏览可用的智能体。 version: 2.5.2 type: procedural risk_level: low status: enabled disable-model-invocation: true tags: - agent - discovery - recommendation metadata: author: desirecore updated_at: '2026-02-28' i18n: default_locale: en-US source_locale: zh-CN locales: - zh-CN - en-US zh-CN: name: 发现智能体 short_desc: 根据需求描述智能推荐最匹配的智能体，引导快速选择 description: 根据用户需求推荐最匹配的智能体，展示候选列表并引导选择。Use when 用户描述需求但不确定该找哪个智能体帮忙，或想浏览可用的智能体。 body: ./SKILL.zh-CN.md source_hash: sha256:28ecd07724adda9a translated_by: human en-US: name: Discover Agent short_desc: Intelligently recommend the best-matching Agent based on the user's needs and guide a quick selection description: >- Recommend the best-matching Agent based on the user's needs, present a candidate list, and guide the user's selection. Use when the user describes a need but is unsure which Agent to ask, or wants to browse available Agents. body: ./SKILL.md source_hash: sha256:28ecd07724adda9a translated_by: ai:claude-opus-4-7 translated_at: '2026-05-03' market: icon: >- category: productivity maintainer: name: DesireCore Official verified: true compatible_agents: [] channel: latest --- # discover-agent skill ## L0: One-Sentence Summary Match and recommend the most suitable registered Agent based on the user's described needs. ## L1: Overview and Use Cases ### Capability Description discover-agent is a **Procedural Skill** that gives DesireCore the ability to discover and recommend a suitable Agent for the user. It understands the user's described needs, performs multi-dimensional matching across the registered Agent list, and presents a candidate list for the user to choose from. ### Use Cases - The user describes a need but doesn't know which Agent to ask for help - The user wants to browse currently available Agents and their capabilities - The user needs to find the best specialist assistant for a specific task - A new user trying the system for the first time needs to learn which Agents are available ### Core Value - **Lower the barrier**: Users don't have to remember each Agent's name and capabilities - **Precise matching**: Smart recommendations based on the semantics of the need, not simple keyword search - **Smooth handoff**: When there's no match, automatically suggest creating a new Agent (handing off to the create-agent skill) ## L2: Detailed Specification ### Execution Flow ``` ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │ 需求理解 │ ──→ │ Agent 检索 │ ──→ │ 匹配评分 │ └──────────────┘ └──────────────┘ └──────────────┘ │ ↓ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │ 引导选择 │ ←── │ 结果展示 │ ←── │ 候选排序 │ └──────────────┘ └──────────────┘ └──────────────┘ ``` ### Stage 1: Need Understanding **Trigger conditions** (any one is sufficient): - The user says "find me a...", "is there a...", "who can help me..." - The user describes a task without specifying a particular Agent - The user says "what Agents are there?", "show me who's available" - The system detects that the user's need does not match the current Agent's capabilities **Need parsing**: Extract the following dimensions from the user's description: | Dimension | Description | Examples | | ----------- | ----------------- | --------------------------------------- | | `domain` | Specialty domain | Legal, finance, technology, education | | `task_type` | Task type | Consultation, review, analysis, writing | | `keywords` | Keywords | Contract, report, code, paper | | `urgency` | Urgency level | Routine / urgent | ### Stage 2: Agent Retrieval **Data source**: Call `GET /api/agents` to fetch the list of all registered Agents. **API call**: ```bash GET /api/agents ``` **Key fields in the returned data**: - `id` — Unique Agent identifier - `name` — Agent name - `description` — Agent description - `skills` — Skill list - `status` — Current status (online/offline/busy) **Filter rules**: - By default, show only Agents with `status: online` or `status: offline` - Exclude system-internal Agents (e.g. DesireCore itself, unless the user explicitly requests them) ### Stage 3: Match Evaluation Comprehensively judge match degree based on the following dimensions (using LLM semantic understanding rather than formula-based computation): | Dimension | Description | | ---------------------- | ------------------------------------------------------------------------ | | Description relevance | Semantic relevance between Agent description / persona and user's need | | Skill match | Relevance of the Agent's skills to the task type | | Domain fit | Fit between the Agent's specialty domain and the user's need domain | | Status availability | Agent's current status (online preferred over offline) | **Display rules**: - Highly matched (clearly suited to the task) → tag as "Recommended" - Partial match (may help) → tag as "Possibly relevant" - No clear relation → do not display ### Stage 4: Candidate Ranking **Ranking rules**: 1. Sort by overall score in descending order 2. On ties, online status takes precedence 3. Show at most 5 candidates ### Stage 5: Result Display **When matches are found**: ``` 根据你的需求，我推荐以下智能体： ┌─────────────────────────────────────────────────────┐ │ 1. 法律顾问助手 匹配度: 92% │ │ 专注合同审查和法律风险评估 │ │ 技能：合同审查、风险评估、法律研究 │ │ 状态：在线 │ ├─────────────────────────────────────────────────────┤ │ 2. AI 文书助手 匹配度: 71% │ │ 专业文书撰写和格式优化 │ │ 技能：文书撰写、格式排版、合规检查 │ │ 状态：在线 │ ├─────────────────────────────────────────────────────┤ │ 3. 数据分析师 匹配度: 45% │ │ 数据分析和可视化报告 │ │ 技能：数据分析、报表生成、趋势预测 │ │ 状态：离线 │ └─────────────────────────────────────────────────────┘ 请选择一个智能体，或告诉我更具体的需求。 ``` **When no matches are found**: ``` 目前没有找到完全匹配你需求的智能体。 你可以： 1. 用更具体的描述再试一次 2. 创建一个新的专业智能体（我可以帮你） 3. 浏览所有可用的智能体 你想怎么做？ ``` **Browse mode** (when the user asks to view all): ``` 当前可用的智能体： 在线： - 法律顾问助手 — 合同审查和法律风险评估 - AI 文书助手 — 专业文书撰写和格式优化 离线： - 数据分析师 — 数据分析和可视化报告 - 翻译助手 — 多语言翻译和本地化 共 4 个智能体。需要了解某个智能体的详细信息吗？ ``` ### Stage 6: Guided Selection **Actions after the user selects**: | User Choice | Follow-up Action | | ------------------------ | ------------------------------------------------------------------ | | Selected an Agent | Switch to a conversation with that Agent and pass the need context | | Asked to learn more | Call `GET /api/agents/:id` for details and present structured info (see below) | | Not satisfied with candidates | Guide the user to refine the need or suggest creating a new Agent | | Chose "create new" | Invoke the create-agent skill, passing the need info already collected | **Implementation of "learn more"**: Call `GET /api/agents/:id` for details, and optionally call structured endpoints for persona/principles: ```bash # 获取基本信息 GET /api/agents/{agentId} # 返回: { id, name, description, skillsCount, toolsCount, status, config, persona, principles } # 获取结构化 persona（可选，用于展示更丰富的信息） GET /api/agents/{agentId}/persona # 返回: { L0, L1: { role, personality, communication_style }, L2 } ``` When presenting to the user, render key information in natural language / table form: ``` 「法律顾问助手」详细信息 | 字段 | 内容 | |------|------| | 角色定位 | 专注合同审查和法律风险评估 | | 性格特征 | 专业、严谨、审慎 | | 技能数量 | 3 个 | | 当前状态 | 在线 | 需要与这个智能体对话吗？ ``` **Context handoff on switch**: ```yaml context_handoff: source_agent: desirecore target_agent: legal-assistant user_intent: '帮我审查这份合同的风险点' ``` ### Collaboration with Other Skills | Collaborating Skill | How It Collaborates | | ------------------- | ----------------------------------------------------------- | | create-agent | When there's no match, suggest creating a new Agent and pass the user's need as initial info | | task-management | After a successful match, optionally auto-create a task and assign it to the target Agent | ### Error Handling | Error Scenario | Handling | | ---------------------------- | --------------------------------------------------------- | | API call fails | Indicate a network error and suggest retrying later | | Agent list is empty | Guide the user to create their first Agent | | User description too vague | Ask follow-up questions and offer domain options to guide | | Recommended Agent has bad status | Mark the status and suggest selecting another online Agent | ### Permission Requirements - Prefer accessing the Agent Service HTTP API via the `Bash` tool with curl - The API base URL is already injected into the "Local API" section of the system prompt; reference it directly - Read-only operations; no risk ### Dependencies - Agent Service HTTP API (`GET /api/agents`) - The local API URL declaration in the system prompt