@yugoo-design/dsh-design-skills-web-app 0.2.7 → 0.2.11

package/README.md

@@ -7,8 +7,11 @@ validators, panel declarations and resources live in `bundles/<name>/`.
 `skill.yaml` remains each Bundle's identity and execution contract. The npm
 package is a distribution container, not the Kit or Bundle lifecycle authority.
 
-`package.json.designSystem.kit` declares the Kit root, Mindset, mount manifest
-and Bundle search root. Load the package root through the generic Kit resolver;
+`package.json.designSystem.kit` declares the built `lib/kits/web_app/` root,
+Mindset, mount manifest and `lib/bundles/` search root. `node build.mjs` copies
+the complete authoring trees and builds the three Flow CLI entries into `lib/`.
+The npm files list selects this complete output; methods, schemas, examples and
+other resources stay inside their owning Kit or Bundle. Load the package root through the generic Kit resolver;
 a bare Kit directory needs its exact Bundle resolver supplied separately.
 Loading content never installs npm dependencies. Repository-only checks and
 preview tools live in `test/` and `scripts/` and are excluded from package files.
@@ -17,28 +20,33 @@ preview tools live in `test/` and `scripts/` and are excluded from package files
 
 | Package | Used by | Purpose |
 | --- | --- | --- |
-| `bpmn-moddle` ^9.0.1 | `bundles/flow_design/scripts/layout.mjs`, `bpmn-validator.mjs` | BPMN XML parse/serialize |
+| `bpmn-moddle` ^9.0.1 | `bundles/flow_design/scripts/layout.mjs`, `layout-metrics.mjs`, `layout-search.mjs` | BPMN XML parse/serialize |
 | `elkjs` ^0.10.0 | `bundles/flow_design/scripts/layout.mjs` | layered graph layout geometry |
 | `bpmn-js` ^18.24.0 | `bundles/web_app_prd_assembly/scripts/visual-renderer.mjs` | conditional BPMN snapshots |
 | `playwright` ^1.61.1 | `bundles/web_app_prd_assembly/scripts/visual-renderer.mjs` | conditional browser snapshots |
 
 These dependencies belong to this engineering package. Source development uses
-the repository's locked dependencies. Copying a bare Bundle does not install its
-script dependencies or provide a browser. A distributable Bundle must include
-or otherwise explicitly provision its required runtime and license materials;
-the loader does not fetch them. Existing dependency diagnostics mention the
-legacy “Kit package.json”; that means this engineering package, not `kit.yaml`.
-Missing dependencies fail with a non-zero exit rather than fabricated output.
-
-## Delivery context 2: packaged Companion payload
-
-In the packaged payload, Kit files live in `kits/web_app/` and Bundle files
-live independently in `bundles/<name>/`. `bundles/flow_design/scripts/layout.mjs` resolves against a
-bundled dist (ELK and bpmn-moddle included); the layout entry was verified
-self-contained by running it outside the repository with empty `NODE_PATH` and
-`NODE_OPTIONS` and no `node_modules` present: exit 0 with BPMN DI injection.
-Bundled third-party code retains its license materials in the distribution
-package (`licenses/`); see `THIRD_PARTY_NOTICES.md` for the Kit-level inventory.
+the repository's locked dependencies. Distribution builds include the Flow
+CLI dependency closure, so a complete distributed Flow Bundle can be copied
+into a project and run without `node_modules`, `NODE_PATH` or an install step.
+The structure/final validator already uses only Node built-ins and is copied
+unchanged. A raw source Bundle still needs the repository's development
+dependencies; loading content never installs them.
+
+The npm content package and both Companion packaging routes reuse the same
+three independent Node 22 ESM entries: layout, metrics and search. Search keeps
+its explicit sibling metrics import to preserve both CLI entry guards. Matching
+original licenses, notices and existing ELK source-availability materials live
+inside `flow_design/licenses/`; project LICENSE, NOTICE and inventory texts also
+remain at the Flow root, so copying the complete Bundle retains them.
+`flow-build.json` records the actual observed build inputs and original-material
+identities; it does not claim a complete nested-prebundle inventory or license
+acceptance. The existing normal-build module observations bind output bytes and
+source inputs for DSH release preparation.
+
+`node --test test/portable-distribution.mjs` builds and packs the content package,
+extracts its full Flow Bundle outside the repository, and exercises layout,
+metrics, search and structure/final validation with no ancestor `node_modules`.
 
 ## flow_design layout contract (verified)
 
@@ -92,12 +100,11 @@ Retention mapping for the user-authorized detail simplification:
   not rewritten. Native-reference implementations are
   retained; this change targets the official declarative Panel, not old renderers.
 
-The 1.0 decision above records the internal Function detail update. The Flow
-execution-resource correction declares `flow_design@1.1`, mounted by
-`web_app@1.1`, for its next content release; the other 14 Bundles retain 1.0.
-Existing project pins and retained snapshots stay unchanged. Adopting new
-content requires an explicit upgrade through a Host with verified native
-migration isolation; a plugin update alone does not upgrade a project.
+The 1.0 decision above records the internal Function detail update. Flow and
+Kit version metadata is upstream provenance. Each project owns its complete
+saved definitions; a plugin update does not replace those files. An Agent may
+compare and revise an explicitly selected project copy through the existing
+editing and validation path, preserving user content without a version gate.
 
 Verification is reported separately for conversion/validation, rendered edits,
 durable save/reopen and native-model consumption; these claims are not implied
@@ -116,8 +123,19 @@ Retention mapping for the Mindset reorganization:
 - The complete experience guidance moves from the opening to section 5.
 - Case strategies and enterprise calibration move to sections 6–7; reconstruction
   and source-evidence review remain complete in section 8.
-- The complete design_skills catalog moves to the end, with identical entries.
-  It remains persisted runtime input, not merely a hidden UI representation.
+- The ordered `design_skills` headings remain at the end as the runtime
+  membership list, matching `kit.yaml`. Capability descriptions come from each
+  mounted Bundle's `SKILL.md` frontmatter rather than repeating them in Mindset.
+  The professional guidance in sections 1–8 remains complete.
+- Those descriptions retain the same-source AI graph and Host boundary, Plugin
+  URL/Secret requirements/Prompt and recovery, observable acceptance scenarios
+  and grounded measurable quality targets, and shared whiteboard scene files.
+  Existing applicability and negative routing conditions remain intact.
+- The AI Bundle's current independent Agent/resource records and same-source
+  graph supersede the old catalog's grouped-card and supporting-harness wording.
+  Its existing method does not require manual functional groups or SDK selection.
+  Saving a design does not execute its tools or reconfigure the coding host;
+  no new behavior or field is introduced.
 - A purpose introduction and AI-capability candidate row clarify existing scope.
 - “Producers own product design choices” becomes “selected Bundle executor” to
   include the existing main execution mode (architecture baseline, Execution).

package/{bundles → lib/bundles}/ai_agent_capabilities/SKILL.md

@@ -4,8 +4,8 @@ description: |-
   - **Capability**: design the AI behavior embedded in a Web App as Agent or direct-model units, with user experience, complete prompts, editable Skills and tools/MCP, and Agent collaboration.
   - **Resolves**: unclear responsibilities, triggers and results, action-condition-result rules, exceptions, materials and long-term memory, and per-Agent handoffs.
   - **Candidate when**: AI behavior needs to be designed, reconstructed from evidence, or revised; required Skills and tool capabilities can be described before selecting an implementation.
-  - **Do not infer**: a chat box, dependency or model name alone establishes the needed mechanisms. Saving the design does not install resources, connect services or prove evaluation success.
-  - **Output**: an editable AI design panel with linked Agents, Skills and tools; preserve complete instructions, user configuration and confirmation choices.
+  - **Do not infer**: a chat box, dependency or model name alone establishes the needed mechanisms. Saving the design does not install resources, connect services, execute tools, configure the coding host or prove evaluation success.
+  - **Output**: an editable AI design panel with linked Agents, Skills and tools; its collaboration graph opens the same canonical records. Preserve complete instructions, user configuration and confirmation choices.
 ---
 
 # AI Agent Design

package/{bundles → lib/bundles}/ai_agent_capabilities/locales/en-US.json

@@ -2,8 +2,8 @@
   "bundleDescription": "- **Capability**: design the AI behavior embedded in a Web App as Agent or direct-model units, with user experience, complete prompts, editable Skills and tools/MCP, and Agent collaboration.\n- **Resolves**: unclear responsibilities, triggers and results, action-condition-result rules, exceptions, materials and long-term memory, and per-Agent handoffs.\n- **Candidate when**: AI behavior needs to be designed, reconstructed from evidence, or revised; required Skills and tool capabilities can be described before selecting an implementation.\n- **Do not infer**: a chat box, dependency or model name alone establishes the needed mechanisms. Saving the design does not install resources, connect services or prove evaluation success.\n- **Output**: an editable AI design panel with linked Agents, Skills and tools; preserve complete instructions, user configuration and confirmation choices.",
   "agents": "AI design units",
   "agent": "Unit",
-  "newAgent": "New Agent",
-  "numberedAgent": "New Agent {index}",
+  "newAgent": "New agent",
+  "numberedAgent": "New agent {index}",
   "empty": "Add an Agent and reference Skills and tools as needed.",
   "guidance": "Check whether the Agent goals, responsibilities and tools fit together, and whether capabilities are missing or unnecessary. Saving here records the design; it does not install or run tools.",
   "name": "Name",
@@ -14,9 +14,9 @@
   "basics": "Goal and outcome",
   "result": "Expected result",
   "boundary": "What must not happen",
-  "entries": "Entries",
-  "entry": "Entry",
-  "entryName": "Entry name",
+  "entries": "Entry points",
+  "entry": "Entry point",
+  "entryName": "Entry point name",
   "trigger": "Trigger method",
   "conversation": "Conversation",
   "action": "Button or menu",
@@ -27,7 +27,7 @@
   "actionLocation": "Action entry location",
   "backgroundCondition": "When it runs",
   "conversationInputs": "Supported input methods",
-  "actionInputs": "Additional user input (optional)",
+  "actionInputs": "Additional user input",
   "autoContext": "Automatically included information",
   "autoContextPlaceholder": "For example, selected text, the current document or the triggering event's content. Leave blank if none.",
   "feedback": "Progress feedback",
@@ -36,7 +36,7 @@
   "voice": "Audio / voice",
   "vision": "Images",
   "selection": "Selected content",
-  "entryHelp": "A unit can have multiple entries. Upload controls do not establish parsing, realtime voice or vision support.",
+  "entryHelp": "A design unit can have multiple entry points. File, voice and image controls alone do not mean the model can parse files, support real-time voice or understand images.",
   "instructions": "Working instructions",
   "systemPrompt": "System prompt",
   "promptHelp": "Complete editable instructions: responsibility, evidence, method, output and exit.",
@@ -49,11 +49,11 @@
   "capabilityPlaceholder": "Describe the ability needed, the information it uses and the useful result it returns.",
   "activation": "When to use",
   "skillBody": "Complete Skill body",
-  "source": "Source (optional)",
+  "source": "Source",
   "sourcePlaceholder": "GitHub URL, local path or provider; leave blank for original content.",
-  "resourceFiles": "Supporting files (optional)",
+  "resourceFiles": "Supporting files",
   "resourceFilesHelp": "List the locations of documents, templates or scripts used by the body. This list does not upload files.",
-  "connectionDetails": "Connection details (optional)",
+  "connectionDetails": "Connection details",
   "resourceEvidence": "Source, example or unknowns",
   "mcp": "MCP connections",
   "mcpSlot": "MCP",
@@ -101,7 +101,7 @@
   "agentRoot": "AI Agents",
   "skillRoot": "Skills",
   "toolRoot": "Tools / MCP",
-  "newskillRoot": "New Skill",
+  "newskillRoot": "New skill",
   "newtoolRoot": "New tool",
   "resourceRefs": "Available methods and capabilities",
   "resourceOrigin": "Source and references",
@@ -113,7 +113,7 @@
   "toolInterface": "Tool description and input/output",
   "connection": "Configuration",
   "toolConnectionHelp": "Keep the function interface or complete MCP connection configuration. Designing does not connect or invoke it.",
-  "authority": "Usage limits (optional)",
+  "authority": "Usage limits",
   "collaboration": "Agent collaboration",
   "delegateRefs": "Other Agents this unit can call",
   "handoff": "Responsibilities and handoff",
@@ -129,10 +129,10 @@
   "agentDesign": "Agent design",
   "experienceTab": "User experience",
   "instructionsTab": "Prompts and resources",
-  "behaviorTab": "Behavior rules",
+  "behaviorTab": "Behavior and exceptions",
   "resourceUse": "Resource use agreements",
   "scopeGroup": "Availability and boundaries",
-  "implementationDetails": "Implementation reference (optional)",
+  "implementationDetails": "Implementation reference",
   "actions": "Behavior rules",
   "actionName": "Action",
   "actionCondition": "Condition",
@@ -152,25 +152,25 @@
   "modelService": "Configuration plan",
   "modelOtherDescription": "Connection description",
   "modelServicePlaceholder": "Describe the intended service and connection approach, or provide a configuration template. Use placeholders such as ${MODEL_API_KEY}, not real keys. Unknown parameters may remain blank.",
-  "integrationRef": "Reference a system integration (optional)",
+  "integrationRef": "Reference a system integration",
   "integrationRefHelp": "Enter a plan below, or select an existing integration without duplicating its configuration.",
   "exceptionCondition": "Exception",
   "exceptionHandling": "Response",
   "collaborator": "Collaborator",
-  "localCollaborator": "Existing Agent (optional)",
-  "collaboratorHelp": "Choose an existing Agent, or leave this blank for an external collaborator. Clearing the selection restores any retained external target.",
+  "localCollaborator": "Existing agent",
+  "collaboratorHelp": "Choose an existing agent, or leave this blank to enter an external collaborator. Clearing the selection restores any external collaborator you entered earlier.",
   "externalCollaborator": "External collaborator",
   "collaborationCondition": "When to hand off",
-  "collaborationHandoff": "Task and material to hand over",
-  "collaborationResult": "Return and failure handling",
-  "modelSourceRoot": "Model entries",
-  "newModelSource": "New model entry",
+  "collaborationHandoff": "Task and context to pass along",
+  "collaborationResult": "Results and failure handling",
+  "modelSourceRoot": "Model connections",
+  "newModelSource": "New model connection",
   "knowledgeResourceRoot": "Knowledge and memory",
   "newKnowledgeResource": "New knowledge resource",
-  "modelSourceRef": "Shared model entry (optional)",
-  "modelSelection": "Model choice (optional)",
-  "modelSelectionHelp": "An intended model, not verified runtime configuration. Agents can select different models through one entry.",
-  "modelSourceHelp": "Select an established entry. The selected entry owns connection intent; retained Agent connection fields are inactive until you clear the selection.",
+  "modelSourceRef": "Shared model connection",
+  "modelSelection": "Model choice",
+  "modelSelectionHelp": "Record the intended model; this is not a verified runtime configuration. Different agents can use different models through the same connection.",
+  "modelSourceHelp": "Select a defined connection to use its settings. The agent’s individual connection settings are kept but inactive until you clear this selection.",
   "resourceDescription": "Purpose and connection intent",
   "knowledgeResourceRefs": "Knowledge and memory used",
   "knowledgeKind": "Resource type",
@@ -184,10 +184,11 @@
   "memoryFormationUpdateHelp": "Describe user-saved or Agent-curated memory, and whether updates or corrections need user confirmation.",
   "memoryUserManagement": "User control",
   "memoryUserManagementHelp": "Describe how users inspect, correct and clear memory, and what happens after clearing. Keep undecided behavior explicit.",
-  "knowledgeAccessScope": "Access and sharing scope (optional)",
+  "knowledgeAccessScope": "Access and sharing scope",
   "knowledgeAccessScopePlaceholder": "For example: private to a user, current project or shared with a team. Leave blank when no extra requirement is needed.",
   "knowledgeResourceHelp": "Resource records own their content and maintenance plan, not access grants or shared user data. Add only this Agent’s use conditions and differences, not a copied resource definition.",
-  "modelSourceLink": "Model entry",
+  "modelSourceLink": "Model connection",
   "knowledgeResourceLink": "Knowledge / memory",
-  "legacyConnectionGroup": "Per-Agent connection (when no shared entry is selected)"
+  "legacyConnectionGroup": "Individual agent connection (when no shared connection is selected)",
+  "knowledgeResource": "Knowledge resource"
 }

package/{bundles → lib/bundles}/ai_agent_capabilities/locales/zh-CN.json

@@ -2,14 +2,14 @@
   "bundleDescription": "- **能力**：以 Agent 或单次模型调用为单元，设计 Web App 中的 AI 功能，包括使用体验、完整提示词、可编辑的技能与工具/MCP，以及 Agent 协作。\n- **解决**：职责、触发方式和结果不清晰，行为规则中的动作、条件、结果，以及异常处理、资料与长期记忆、Agent 交接缺少设计的问题。\n- **适用时机**：需要设计、基于证据还原或调整 AI 功能时；可以先明确所需技能与工具能力，再选择具体实现。\n- **不要推断**：不能仅凭对话框、依赖或模型名称认定 AI 机制已经明确。保存设计不会安装资源、连接服务，也不代表通过实际评测。\n- **输出**：相互引用的 Agent、技能与工具设计面板，保留完整指令、用户配置和确认选择。",
   "agents": "AI 设计单元",
   "agent": "设计单元",
-  "newAgent": "新 Agent",
-  "numberedAgent": "新 Agent {index}",
-  "empty": "添加一个 Agent，按需引用技能和工具。",
-  "guidance": "核对 Agent 的目标、能力分工和工具是否匹配，有无缺失或不必要的能力；这里保存的是设计，不会安装或执行工具。",
+  "newAgent": "新智能体",
+  "numberedAgent": "新智能体 {index}",
+  "empty": "添加一个智能体，按需引用技能和工具。",
+  "guidance": "核对智能体的目标、能力分工和工具是否匹配，有无缺失或不必要的能力；这里保存的是设计，不会安装或执行工具。",
   "name": "名称",
   "description": "职责与目标",
   "functionRefs": "关联功能",
-  "agentKind": "Agent",
+  "agentKind": "智能体",
   "modelCall": "模型直调",
   "basics": "目标与结果",
   "result": "预期结果",
@@ -27,7 +27,7 @@
   "actionLocation": "操作入口",
   "backgroundCondition": "触发时机",
   "conversationInputs": "支持的输入方式",
-  "actionInputs": "用户补充输入（选填）",
+  "actionInputs": "用户补充输入",
   "autoContext": "自动带入的信息",
   "autoContextPlaceholder": "例如选中的文字、当前文档或触发事件携带的内容；没有则留空。",
   "feedback": "过程反馈",
@@ -40,20 +40,20 @@
   "instructions": "工作说明",
   "systemPrompt": "系统提示词",
   "promptHelp": "完整可编辑的指令正文：职责、证据、方法、输出与退出条件。",
-  "resources": "Skill、MCP 与工具",
+  "resources": "技能、MCP 与工具",
   "resourcesHelp": "选择已定义的资源；完整内容在左侧独立维护。这里补充使用范围。只有用户要求时才搜索推荐，保存不会安装或连接。",
-  "skills": "Skill 配置",
-  "skillSlot": "Skill",
+  "skills": "技能配置",
+  "skillSlot": "技能",
   "slotName": "名称",
   "purpose": "能力说明",
   "capabilityPlaceholder": "说明需要什么能力、会使用什么信息、应给出什么结果。",
   "activation": "适用情形",
-  "skillBody": "完整 Skill 正文",
-  "source": "来源（选填）",
+  "skillBody": "完整技能正文",
+  "source": "来源",
   "sourcePlaceholder": "GitHub 地址、本地路径或提供方；自行编写可留空。",
-  "resourceFiles": "附带文件（可选）",
+  "resourceFiles": "附带文件",
   "resourceFilesHelp": "记录正文会用到的参考文档、模板或脚本的位置；这里维护文件清单，不上传文件。",
-  "connectionDetails": "接入配置（可选）",
+  "connectionDetails": "接入配置",
   "resourceEvidence": "来源、样例或待确认项",
   "mcp": "MCP 配置",
   "mcpSlot": "MCP",
@@ -98,7 +98,7 @@
   "evidence": "设计依据与未确定事项",
   "references": "固定版本来源",
   "other": "已有／自建／其他",
-  "agentRoot": "AI Agent",
+  "agentRoot": "AI 智能体",
   "skillRoot": "技能",
   "toolRoot": "工具 / MCP",
   "newskillRoot": "新技能",
@@ -113,9 +113,9 @@
   "toolInterface": "工具说明与输入输出",
   "connection": "配置内容",
   "toolConnectionHelp": "函数工具保留接口定义；MCP 保留完整连接配置。这里只设计，不调用或连接。",
-  "authority": "使用限制（可选）",
-  "collaboration": "Agent 协作",
-  "delegateRefs": "可以调用的其他 Agent",
+  "authority": "使用限制",
+  "collaboration": "智能体协作",
+  "delegateRefs": "可以调用的其他智能体",
   "handoff": "各自负责什么、如何交接",
   "externalDelegation": "其他协作方（如有）",
   "relationTitle": "AI 资源关系",
@@ -126,13 +126,13 @@
   "modelAccess": "模型使用方式",
   "inheritModelAccess": "沿用已有模型接入",
   "configureModelAccess": "由产品提供模型配置",
-  "agentDesign": "Agent 设计",
+  "agentDesign": "智能体设计",
   "experienceTab": "使用体验",
   "instructionsTab": "提示词与资源",
-  "behaviorTab": "行为规则",
+  "behaviorTab": "行为与异常",
   "resourceUse": "资源使用约定",
   "scopeGroup": "使用条件与边界",
-  "implementationDetails": "实现参考（可选）",
+  "implementationDetails": "实现参考",
   "actions": "行为规则",
   "actionName": "动作",
   "actionCondition": "条件",
@@ -152,25 +152,25 @@
   "modelService": "配置方案",
   "modelOtherDescription": "接入说明",
   "modelServicePlaceholder": "说明拟使用的服务和接入安排，也可填写配置模板；真实密钥请用 ${MODEL_API_KEY} 等占位符。尚未确定的参数可留空。",
-  "integrationRef": "引用系统集成（可选）",
+  "integrationRef": "引用系统集成",
   "integrationRefHelp": "可以直接填写下方配置方案；已有系统集成时选择引用，无需重复填写其配置。",
   "exceptionCondition": "异常情况",
   "exceptionHandling": "处理方式",
   "collaborator": "协作对象",
-  "localCollaborator": "已有 Agent（可选）",
-  "collaboratorHelp": "选择已有 Agent，或留空填写外部协作方。清空选择后可恢复已填的外部对象。",
+  "localCollaborator": "已有智能体",
+  "collaboratorHelp": "选择已有智能体，或留空填写外部协作方。清空选择后可恢复已填的外部对象。",
   "externalCollaborator": "外部协作方",
   "collaborationCondition": "何时交接",
   "collaborationHandoff": "交给它什么",
   "collaborationResult": "返回结果与失败处理",
-  "modelSourceRoot": "模型入口",
-  "newModelSource": "新模型入口",
+  "modelSourceRoot": "模型接入",
+  "newModelSource": "新模型接入",
   "knowledgeResourceRoot": "知识与记忆",
   "newKnowledgeResource": "新知识资源",
-  "modelSourceRef": "共享模型入口（可选）",
-  "modelSelection": "模型选择（可选）",
-  "modelSelectionHelp": "记录期望使用的模型，不是已验证的运行配置。同一入口可以为不同 Agent 选择不同模型。",
-  "modelSourceHelp": "选择已明确的入口后，由入口持有接入意图；原 Agent 接入字段保留但停用，清空选择后恢复。",
+  "modelSourceRef": "共享模型接入",
+  "modelSelection": "模型选择",
+  "modelSelectionHelp": "记录期望使用的模型，不是已验证的运行配置。同一接入方式可以为不同智能体选择不同模型。",
+  "modelSourceHelp": "选择已明确的接入方式后，将使用该方式的配置；原智能体的独立接入配置会保留但停用，清空选择后恢复。",
   "resourceDescription": "用途与接入意图",
   "knowledgeResourceRefs": "使用的知识与记忆",
   "knowledgeKind": "资源类型",
@@ -181,13 +181,14 @@
   "knowledgeSourceMaintenance": "来源与维护",
   "knowledgeSourceMaintenanceHelp": "说明资料来自哪里、由谁更新；内置资料、用户上传或已有业务资料均可，不需要数据库和接口参数。",
   "memoryFormationUpdate": "形成与更新",
-  "memoryFormationUpdateHelp": "说明由用户主动保存还是 Agent 整理，以及更新、纠错是否需要用户确认。",
+  "memoryFormationUpdateHelp": "说明由用户主动保存还是智能体整理，以及更新、纠错是否需要用户确认。",
   "memoryUserManagement": "用户管理",
   "memoryUserManagementHelp": "说明用户如何查看、纠正和清除记忆，以及清除后的行为；尚未确定的保留为待确认项。",
-  "knowledgeAccessScope": "使用与共享范围（选填）",
+  "knowledgeAccessScope": "使用与共享范围",
   "knowledgeAccessScopePlaceholder": "如个人私有、当前项目或团队共享；没有额外要求可以留空。",
-  "knowledgeResourceHelp": "资源条目定义内容与维护方式，不授予访问权限或共享用户数据。这里只补充当前 Agent 的使用条件与差异，不重复整份资源说明。",
-  "modelSourceLink": "模型入口",
+  "knowledgeResourceHelp": "资源条目定义内容与维护方式，不授予访问权限或共享用户数据。这里只补充当前智能体的使用条件与差异，不重复整份资源说明。",
+  "modelSourceLink": "模型接入",
   "knowledgeResourceLink": "知识 / 记忆",
-  "legacyConnectionGroup": "Agent 独立接入（未选择共享入口时）"
+  "legacyConnectionGroup": "智能体独立接入（未选择共享接入方式时）",
+  "knowledgeResource": "知识资源"
 }

package/{bundles → lib/bundles}/ai_agent_capabilities/panel.yaml

@@ -11,7 +11,7 @@ empty_description: $t:empty
 sections:
   - id: skills
     title: $t:skillRoot
-    item_name: $t:skillRoot
+    item_name: $t:skillSlot
     confirm_delete: true
     create:
       id_field: id
@@ -47,7 +47,7 @@ sections:
           field: notes
   - id: tools
     title: $t:toolRoot
-    item_name: $t:toolRoot
+    item_name: $t:tool
     confirm_delete: true
     create:
       id_field: id
@@ -92,7 +92,7 @@ sections:
           field: notes
   - id: model_sources
     title: $t:modelSourceRoot
-    item_name: $t:modelSourceRoot
+    item_name: $t:modelSourceLink
     confirm_delete: true
     create:
       id_field: id
@@ -104,21 +104,21 @@ sections:
         variant: detail
         title: {field: name, editable: true}
       body:
-        - kind: field
-          field: model_connection
-        - kind: field
-          field: integration_ref
         - kind: group
           title: $t:resourceDescription
           items:
+            - kind: field
+              field: model_connection
+            - kind: field
+              field: integration_ref
             - kind: field
               field: description
               label: visually_hidden
-        - kind: field
-          field: notes
+            - kind: field
+              field: notes
   - id: knowledge_resources
     title: $t:knowledgeResourceRoot
-    item_name: $t:knowledgeResourceRoot
+    item_name: $t:knowledgeResource
     confirm_delete: true
     create:
       id_field: id
@@ -165,67 +165,48 @@ sections:
           field: name
           editable: true
       body:
-        - kind: tabs
-          id: agent-design
-          title: $t:agentDesign
-          tabs:
-            - id: experience
-              title: $t:experienceTab
-              items:
-                - kind: field
-                  field: description
-                - kind: field
-                  field: entries
-                - kind: group
-                  title: $t:contextGroup
-                  collapsible: true
-                  items:
-                    - kind: field
-                      field: knowledge_resource_refs
-                    - kind: field
-                      field: context
-                    - kind: field
-                      field: memory_retention
-            - id: instructions
-              title: $t:instructionsTab
-              items:
-                - kind: field
-                  field: system_prompt
-                - kind: field
-                  field: skill_refs
-                - kind: field
-                  field: tool_refs
-            - id: behavior
-              title: $t:behaviorTab
-              items:
-                - kind: field
-                  field: actions
-                - kind: field
-                  field: exceptions
-                - kind: field
-                  field: notes
-            - id: collaboration
-              title: $t:collaboration
-              items:
-                - kind: field
-                  field: collaborations
-            - id: model-connection
-              title: $t:modelConnectionTab
-              items:
-                - kind: field
-                  field: model_source_ref
-                - kind: field
-                  field: model_selection
-                - kind: group
-                  title: $t:legacyConnectionGroup
-                  collapsible: true
-                  items:
-                    - kind: field
-                      field: model_connection
-                    - kind: field
-                      field: integration_ref
-                    - kind: field
-                      field: model_service
+        - {kind: field, field: description}
+        - {kind: field, field: entries}
+        - kind: group
+          title: $t:behaviorTab
+          collapsible: true
+          default_open: true
+          items:
+            - {kind: field, field: actions}
+            - {kind: field, field: exceptions}
+            - {kind: field, field: notes}
+        - kind: group
+          title: $t:contextGroup
+          collapsible: true
+          items:
+            - {kind: field, field: knowledge_resource_refs}
+            - {kind: field, field: context}
+            - {kind: field, field: memory_retention}
+        - kind: group
+          title: $t:collaboration
+          collapsible: true
+          items:
+            - {kind: field, field: collaborations}
+        - kind: group
+          title: $t:instructionsTab
+          collapsible: true
+          items:
+            - {kind: field, field: system_prompt}
+            - {kind: field, field: skill_refs}
+            - {kind: field, field: tool_refs}
+        - kind: group
+          title: $t:modelConnectionTab
+          collapsible: true
+          items:
+            - {kind: field, field: model_source_ref}
+            - {kind: field, field: model_selection}
+        - kind: group
+          title: $t:legacyConnectionGroup
+          collapsible: true
+          items:
+            - {kind: field, field: model_connection}
+            - {kind: field, field: integration_ref}
+            - {kind: field, field: model_service}
 fields:
   - key: model_source_ref
     label: $t:modelSourceRef
@@ -324,6 +305,9 @@ fields:
           - {field: trigger, in: [conversation], label: '$t:conversationLocation'}
           - {field: trigger, in: [action], label: '$t:actionLocation'}
           - {field: trigger, in: [background], label: '$t:backgroundCondition'}
+      - key: feedback
+        label: $t:feedback
+        control: textarea
       - key: input_modes
         label: $t:inputModes
         control: token_list
@@ -331,7 +315,6 @@ fields:
         label_when:
           - {field: trigger, in: [conversation], label: '$t:conversationInputs'}
           - {field: trigger, in: [action], label: '$t:actionInputs'}
-        token_layout: cards
         token_add_mode: menu
         options:
           - value: text
@@ -348,9 +331,6 @@ fields:
         label: $t:autoContext
         control: textarea
         placeholder: $t:autoContextPlaceholder
-      - key: feedback
-        label: $t:feedback
-        control: textarea
     item_label_field: name
     item_defaults:
       name: ""

package/{bundles → lib/bundles}/business_data/locales/en-US.json

@@ -18,6 +18,6 @@
   "source": "Source object (subtype for inheritance)", "target": "Target object (parent for inheritance)",
   "sourceCount": "How many source objects per target object?", "targetCount": "How many target objects per source object?",
   "unknownCount": "Leave blank when undecided. Blank does not mean zero or exactly one.",
-  "countOne": "Exactly one", "countOptional": "None or one", "countMany": "One or more", "countAny": "None or more",
-  "relationMeaning": "Meaning and constraints (optional)", "relationMeaningHint": "Describe uncertainty about quantities, ownership or deletion behavior here."
+  "countOne": "Exactly one", "countOptional": "Zero or one", "countMany": "One or more", "countAny": "Zero or more",
+  "relationMeaning": "Meaning and constraints", "relationMeaningHint": "Describe uncertainty about quantities, ownership or deletion behavior here."
 }

package/{bundles → lib/bundles}/business_data/locales/zh-CN.json

@@ -19,5 +19,5 @@
   "sourceCount": "每个终点对象关联多少个起点对象？", "targetCount": "每个起点对象关联多少个终点对象？",
   "unknownCount": "尚未确定时留空，不代表零个或仅一个。",
   "countOne": "必须且仅一个", "countOptional": "可以没有，最多一个", "countMany": "至少一个，可有多个", "countAny": "可以没有，也可有多个",
-  "relationMeaning": "含义与约束（选填）", "relationMeaningHint": "数量、归属或删除时的约定尚未确定，也可以在这里说明。"
+  "relationMeaning": "含义与约束", "relationMeaningHint": "数量、归属或删除时的约定尚未确定，也可以在这里说明。"
 }