feat: 新增代码智能内置技能 (#61)

## 变更摘要 - 新增中英文 `code-intelligence` 内置技能，按需暴露隐藏的 `Lsp` 工具 - 记录 9 种语义导航操作、目录权限边界、缺少语言服务器时的降级策略 - 要求客户端版本 `10.0.94`，避免旧客户端安装后缺失工具 - 更新内置技能清单、市场统计和版本 ## 验证 - `uv run --with httpx --with ruamel-yaml scripts/i18n/validate-i18n.py` - `uv run --with httpx --with ruamel-yaml scripts/i18n/translate.py --check` ## Summary Adds the bilingual `code-intelligence` built-in skill, exposes the hidden `Lsp` tool only when the skill is active, documents its permission and fallback behavior, and updates the market manifest and counts.
2026-07-23 02:23:48 +08:00 · 2026-07-22 20:03:56 +08:00
parent 29a599d6a9
commit 114f03886b
5 changed files with 206 additions and 6 deletions
--- a/README.md
+++ b/README.md
@@ -23,16 +23,16 @@ DesireCore 官方市场仓库，存放官方维护的 Agent/Skill 定义，以
 The market currently contains:

 - `1` Agent: `desirecore`
- `30` local built-in skills with `SKILL.md`
+- `31` local built-in skills with `SKILL.md`
 - `22` external skill entries with `entry.json`
- `52` publishable skills in total (`SKILL.md` + `entry.json`)
+- `53` publishable skills in total (`SKILL.md` + `entry.json`)

 ## Skill Sources

 Local built-in skills are installable from this repository and must be listed in `builtin-skills.json`:

 ```text
-configuring-compute, create-agent, dashscope-image-gen, delete-agent,
+code-intelligence, configuring-compute, create-agent, dashscope-image-gen, delete-agent,
 dev-environment-setup, discover-agent, docx, frontend-design, guizang-ppt,
 image-to-image, mail-operations, manage-skills, manage-teams, markdown,
 minimax-music-gen, minimax-video-gen, nodejs-runtime, pdf, pptx,
--- a/builtin-skills.json
+++ b/builtin-skills.json
@@ -1,5 +1,6 @@
 {
  "skills": [
+    "code-intelligence",
    "configuring-compute",
    "create-agent",
    "dashscope-image-gen",
--- a/manifest.json
+++ b/manifest.json
@@ -1,6 +1,6 @@
 {
  "name": "DesireCore Official Market",
-  "version": "1.2.15",
+  "version": "1.2.16",
  "schemaVersion": "1.1.0",
  "supportedLocales": ["zh-CN", "en-US"],
  "defaultLocale": "en-US",
@@ -27,8 +27,8 @@
  },
  "stats": {
    "totalAgents": 1,
-    "totalSkills": 52,
-    "lastUpdated": "2026-07-19"
+    "totalSkills": 53,
+    "lastUpdated": "2026-07-22"
  },
  "features": [
    "curated-index",
--- a/skills/code-intelligence/SKILL.md
+++ b/skills/code-intelligence/SKILL.md
@@ -0,0 +1,133 @@
+---
+name: code-intelligence
+description: >-
+  Use this skill when the user needs semantic code navigation backed by a
+  Language Server: jump to definitions or implementations, find references,
+  inspect hover/type information, list document or workspace symbols, or trace
+  incoming and outgoing calls. Activate it for requests such as "where is this
+  symbol defined", "find usages", "who calls this function", "show the file
+  outline", or "search symbols in the workspace". A compatible Language Server
+  must already be installed; this skill never installs one automatically. Use
+  when 用户提到 跳转定义、查找引用、查找实现、类型信息、悬停信息、符号大纲、
+  工作区符号、调用关系、谁调用了这个函数、这个函数调用了谁、语义代码导航。
+version: 1.0.0
+type: procedural
+risk_level: low
+status: enabled
+disable-model-invocation: true
+tags:
+  - development
+  - lsp
+  - code-navigation
+  - symbols
+  - references
+provides:
+  tools:
+    - Lsp
+metadata:
+  author: desirecore
+  updated_at: '2026-07-22'
+  i18n:
+    default_locale: en-US
+    source_locale: zh-CN
+    locales:
+      - zh-CN
+      - en-US
+    zh-CN:
+      name: 代码智能
+      short_desc: 基于 Language Server 的定义、引用、符号与调用关系导航
+      description: >-
+        使用已安装的 Language Server 执行语义代码导航，包括定义、引用、实现、悬停类型、符号大纲和调用层级。
+      body: ./SKILL.zh-CN.md
+      source_hash: sha256:cf613ab2572810e5
+      translated_by: human
+    en-US:
+      name: Code Intelligence
+      short_desc: Language Server powered definitions, references, symbols, and call navigation
+      description: >-
+        Use an installed Language Server for semantic code navigation, including definitions, references, implementations, hover types, symbols, and call hierarchies.
+      body: ./SKILL.md
+      source_hash: sha256:cf613ab2572810e5
+      translated_by: human
+market:
+  icon: >-
+    <svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0
+    24 24" fill="none"><defs><linearGradient id="ci-a" x1="3" y1="3" x2="21"
+    y2="21" gradientUnits="userSpaceOnUse"><stop stop-color="#007AFF"/><stop
+    offset="1" stop-color="#34C759"/></linearGradient></defs><rect x="3" y="3"
+    width="18" height="18" rx="3" fill="url(#ci-a)" fill-opacity="0.1"
+    stroke="url(#ci-a)" stroke-width="1.5"/><path d="M9 8l-3 4 3 4M15
+    8l3 4-3 4M13 6l-2 12" stroke="url(#ci-a)" stroke-width="1.5"
+    stroke-linecap="round" stroke-linejoin="round"/></svg>
+  category: development
+  maintainer:
+    name: DesireCore Official
+    verified: true
+  compatible_agents: []
+  channel: latest
+  required_client_version: 10.0.94
+---
+
+# code-intelligence Skill
+
+## L0: One-line Summary
+
+Use DesireCore's hidden `Lsp` tool for semantic code navigation through an already-installed Language Server.
+
+## L1: When to Use
+
+Activate this skill when the task depends on symbol meaning rather than text matching:
+
+- Jump to a symbol's definition or implementation.
+- Find all semantic references to a symbol.
+- Inspect type information or documentation at a source position.
+- List the symbols in one file or search symbols across the workspace.
+- Find callers and callees through the LSP call hierarchy.
+
+Use `Grep` or `Glob` instead when the task is purely textual or no compatible Language Server is installed.
+
+## L2: Operating Procedure
+
+### 1. Check the workspace boundary
+
+`Lsp` starts an external indexing process that can read the authorized workspace. The target file therefore needs directory-level read access. If the tool reports that an exact-file grant is insufficient, ask the user to authorize the project directory through `ManageWorkDirs`; never widen the scope silently.
+
+### 2. Choose one operation
+
+| Operation | Purpose | Required input |
+|---|---|---|
+| `goToDefinition` | Find where a symbol is defined | file, line, character |
+| `findReferences` | Find semantic references, including the declaration | file, line, character |
+| `hover` | Read type information or documentation | file, line, character |
+| `documentSymbol` | Outline symbols in one file | file |
+| `workspaceSymbol` | Search named symbols in the current workspace | file, non-empty query |
+| `goToImplementation` | Find concrete implementations | file, line, character |
+| `prepareCallHierarchy` | Resolve a callable item at a position | file, line, character |
+| `incomingCalls` | Find functions or methods that call the target | file, line, character |
+| `outgoingCalls` | Find functions or methods called by the target | file, line, character |
+
+`line` and `character` are 1-based, matching editor coordinates. Character offsets use UTF-16 semantics.
+
+### 3. Treat results as bounded navigation evidence
+
+- Results are filtered by the active read scope and workspace `.gitignore` rules.
+- Use `maxResults` to keep broad reference or symbol searches focused; the maximum is 200.
+- A first request may take longer while the server indexes the workspace.
+- Re-run the request after source files change; DesireCore synchronizes the latest saved UTF-8 content.
+
+### 4. Handle missing servers without modifying the system
+
+DesireCore only discovers installed binaries and returns an installation hint when one is missing. Do not run that installation command unless the user explicitly asks you to install the dependency.
+
+Supported built-in mappings:
+
+- TypeScript/JavaScript: `typescript-language-server`
+- Python: `pyright-langserver`
+- Go: `gopls`
+- Rust: `rust-analyzer`
+
+### 5. Fallbacks
+
+- Unsupported file type or missing server: use `Grep`, `Glob`, and `Read` for text-level investigation.
+- Server does not advertise an operation: report that capability mismatch instead of guessing.
+- Empty semantic result: explain that the symbol may be unresolved, excluded, or filtered by access rules; do not claim the symbol has no usages without qualification.
--- a/skills/code-intelligence/SKILL.zh-CN.md
+++ b/skills/code-intelligence/SKILL.zh-CN.md
@@ -0,0 +1,66 @@
+<!-- locale: zh-CN -->
+
+# code-intelligence 技能
+
+## L0：一句话摘要
+
+通过用户已经安装的 Language Server，使用 DesireCore 默认隐藏的 `Lsp` 工具完成语义代码导航。
+
+## L1：何时使用
+
+当任务依赖符号语义而不是文本匹配时激活本技能：
+
+- 跳转到符号的定义或具体实现。
+- 查找符号的全部语义引用。
+- 查看源码位置上的类型信息或文档。
+- 列出单个文件的符号大纲，或在工作区内搜索符号。
+- 通过 LSP 调用层级查找调用者与被调用者。
+
+如果需求只是文本搜索，或者没有兼容的 Language Server，改用 `Grep` 或 `Glob`。
+
+## L2：操作流程
+
+### 1. 检查工作区权限边界
+
+`Lsp` 会启动能够读取授权工作区的外部索引进程，因此目标文件必须具备目录级读取权限。如果工具提示精确文件授权不足，应让用户通过 `ManageWorkDirs` 授权项目目录；不得静默扩大访问范围。
+
+### 2. 选择一个 operation
+
+| Operation | 用途 | 必需输入 |
+|---|---|---|
+| `goToDefinition` | 查找符号定义 | 文件、行、列 |
+| `findReferences` | 查找语义引用，包含声明 | 文件、行、列 |
+| `hover` | 查看类型信息或文档 | 文件、行、列 |
+| `documentSymbol` | 获取单个文件的符号大纲 | 文件 |
+| `workspaceSymbol` | 在当前工作区搜索具名符号 | 文件、非空 query |
+| `goToImplementation` | 查找具体实现 | 文件、行、列 |
+| `prepareCallHierarchy` | 解析当前位置的可调用符号 | 文件、行、列 |
+| `incomingCalls` | 查找谁调用了目标符号 | 文件、行、列 |
+| `outgoingCalls` | 查找目标符号调用了谁 | 文件、行、列 |
+
+`line` 和 `character` 都是与编辑器一致的 1-based 坐标；列偏移使用 UTF-16 语义。
+
+### 3. 把结果作为有边界的导航证据
+
+- 返回位置会经过当前读取范围和工作区 `.gitignore` 过滤。
+- 引用或符号过多时使用 `maxResults` 收敛结果，最大值为 200。
+- 首次调用时 Language Server 可能需要索引工作区，因此耗时会更长。
+- 源码文件变化后重新调用；DesireCore 会同步最新的已保存 UTF-8 内容。
+
+### 4. 缺少 Server 时不得擅自修改系统
+
+DesireCore 只探测用户已经安装的二进制，缺失时返回安装提示。除非用户明确要求安装依赖，否则不得执行该安装命令。
+
+内置支持映射：
+
+- TypeScript/JavaScript：`typescript-language-server`
+- Python：`pyright-langserver`
+- Go：`gopls`
+- Rust：`rust-analyzer`
+
+### 5. 降级策略
+
+- 文件类型不支持或 Server 未安装：使用 `Grep`、`Glob`、`Read` 做文本级调查。
+- Server 未声明某项能力：如实报告 capability 不匹配，不得猜测结果。
+- 语义结果为空：说明符号可能未解析、被排除或被访问规则过滤；不能直接断言“没有引用”。
+