feat(collaboration): complete agent onboarding and capability execution

.github/scripts/classify-wework-desktop-e2e.sh

@@ -7,6 +7,8 @@ core_segments=(
   workspace-tabs
   collaboration-shared-core
   collaboration-settings-matrix
+  collaboration-first-use
+  collaboration-local-agent-capabilities
   collaboration-agent-automation-chain
   cloud-space-mention
   priority-filter
@@ -145,7 +147,7 @@ core_shards=(
   claude-runtime,workspace-tabs,task-attachments
   task-status-sync,task-board-association,core-task-flow,change-request-status,context-compaction
   window-lifecycle,browser-toolbar-actions,browser-annotation-anchors
-  project-automation,collaboration-agent-automation-chain
+  project-automation,collaboration-first-use,collaboration-local-agent-capabilities,collaboration-agent-automation-chain
   resilience,environment-panel-scroll
   workspace-attachments,automation-lifecycle
   project-assignment-notification,split-workbench,priority-filter,project-event-sources,board-focus-view
@@ -522,6 +524,14 @@ classify_wework_path() {
       select_target "core:collaboration-settings-matrix"
       return
       ;;
+    wework/e2e/desktop/scenarios/collaboration-first-use.scenario.mjs)
+      select_target "core:collaboration-first-use"
+      return
+      ;;
+    wework/e2e/desktop/scenarios/collaboration-local-agent-capabilities.scenario.mjs)
+      select_target "core:collaboration-local-agent-capabilities"
+      return
+      ;;
     wework/e2e/desktop/scenarios/collaboration-agent-automation-chain.scenario.mjs)
       select_target "core:collaboration-agent-automation-chain"
       return
@@ -551,6 +561,13 @@ classify_wework_path() {
       wework/src/features/todo/WorkItemComposerGuide*)
       select_target "core:task-status-sync"
       select_target "core:task-board-association"
+      if [[ "$path" == wework/src/api/local/localDelivery* ]]; then
+        select_target "core:collaboration-local-agent-capabilities"
+      fi
+      if [[ "$path" == wework/src/features/todo/CloudTodoWorkspace* || \
+        "$path" == wework/src/features/todo/WorkItemComposerGuide* ]]; then
+        select_target "core:collaboration-first-use"
+      fi
       if [[ "$path" == wework/src/components/layout/useWorkbenchCloudProjectContext* ]]; then
         select_target "core:cloud-context-resilience"
       fi
@@ -927,12 +944,16 @@ classify_path() {
       select_target "core:remote-device-onboarding"
       select_target "core:collaboration-shared-core"
       select_target "core:collaboration-settings-matrix"
+      select_target "core:collaboration-first-use"
+      select_target "core:collaboration-local-agent-capabilities"
       select_target "core:collaboration-agent-automation-chain"
       select_target "cloud:cloud-device-lifecycle"
       ;;
     packages/collaboration/*)
       select_target "core:collaboration-shared-core"
       select_target "core:collaboration-settings-matrix"
+      select_target "core:collaboration-first-use"
+      select_target "core:collaboration-local-agent-capabilities"
       select_target "core:collaboration-agent-automation-chain"
       ;;
     executor/* | packages/chat-core/* | package.json | pnpm-lock.yaml | pnpm-workspace.yaml)

.github/scripts/test-ci-cache-policy.sh

@@ -359,6 +359,16 @@ fi
 if ! sed -n '/^  executor-e2e-tests:/,/^  merge-reports:/p' \
   "$workflow_dir/e2e-tests.yml" |
   grep -F 'prepare-platform-e2e-image' >/dev/null ||
+  ! sed -n '/^  executor-e2e-tests:/,/^  merge-reports:/p' \
+    "$workflow_dir/e2e-tests.yml" |
+    grep -F 'node scripts/prepare-codex-binary.mjs --materialize' >/dev/null ||
+  ! sed -n '/^  executor-e2e-tests:/,/^  merge-reports:/p' \
+    "$workflow_dir/e2e-tests.yml" |
+    grep -F 'CODEX_BINARY_PATH: ${{ github.workspace }}/wework/resources/binaries/codex/x86_64-unknown-linux-gnu/vendor/x86_64-unknown-linux-musl/bin/codex' \
+      >/dev/null ||
+  ! sed -n '/^  executor-e2e-tests:/,/^  merge-reports:/p' \
+    "$workflow_dir/e2e-tests.yml" |
+    grep -F 'test -x "$CODEX_BINARY_PATH"' >/dev/null ||
   ! sed -n '/^  executor-e2e-tests:/,/^  merge-reports:/p' \
     "$workflow_dir/e2e-tests.yml" |
     grep -F \

.github/scripts/test-classify-ci-changes.sh

@@ -430,15 +430,15 @@ wework_desktop_other_e2e_matrix={"include":[]}' \
 assert_desktop_case "shared collaboration package selects its desktop checkpoint" \
   'wework_desktop_e2e=true
 wework_desktop_core_e2e=true
-wework_desktop_core_e2e_matrix={"include":[{"id":"core-5","name":"Core / shard 5","segments":"collaboration-shared-core,collaboration-settings-matrix"},{"id":"core-9","name":"Core / shard 9","segments":"collaboration-agent-automation-chain"}]}
+wework_desktop_core_e2e_matrix={"include":[{"id":"core-5","name":"Core / shard 5","segments":"collaboration-shared-core,collaboration-settings-matrix"},{"id":"core-9","name":"Core / shard 9","segments":"collaboration-first-use,collaboration-local-agent-capabilities,collaboration-agent-automation-chain"}]}
 wework_desktop_other_e2e=false
 wework_desktop_other_e2e_matrix={"include":[]}' \
   "packages/collaboration/src/CollaborationApp.tsx"
 
 assert_desktop_case "collaboration execution environment changes select onboarding and lifecycle coverage" \
   'wework_desktop_e2e=true
 wework_desktop_core_e2e=true
-wework_desktop_core_e2e_matrix={"include":[{"id":"core-2","name":"Core / shard 2","segments":"remote-device-onboarding"},{"id":"core-5","name":"Core / shard 5","segments":"collaboration-shared-core,collaboration-settings-matrix"},{"id":"core-9","name":"Core / shard 9","segments":"collaboration-agent-automation-chain"}]}
+wework_desktop_core_e2e_matrix={"include":[{"id":"core-2","name":"Core / shard 2","segments":"remote-device-onboarding"},{"id":"core-5","name":"Core / shard 5","segments":"collaboration-shared-core,collaboration-settings-matrix"},{"id":"core-9","name":"Core / shard 9","segments":"collaboration-first-use,collaboration-local-agent-capabilities,collaboration-agent-automation-chain"}]}
 wework_desktop_cloud_e2e=true
 wework_desktop_cloud_e2e_matrix={"include":[{"id":"cloud-13","name":"Cloud / shard 13","segments":"cloud-device-lifecycle"}]}
 wework_desktop_other_e2e=false
@@ -575,7 +575,7 @@ wework_desktop_other_e2e=true
 wework_desktop_other_e2e_matrix={"include":[{"id":"plugins","name":"Plugins","command":"e2e:desktop:plugins","segment":""}]}
 wework_desktop_macos_inspector_e2e=true'
 full_desktop_expected="${full_desktop_expected/\"segments\":\"resilience\"/\"segments\":\"resilience,environment-panel-scroll\"}"
-full_desktop_expected="${full_desktop_expected/\"segments\":\"project-automation\"/\"segments\":\"project-automation,collaboration-agent-automation-chain\"}"
+full_desktop_expected="${full_desktop_expected/\"segments\":\"project-automation\"/\"segments\":\"project-automation,collaboration-first-use,collaboration-local-agent-capabilities,collaboration-agent-automation-chain\"}"
 full_desktop_expected="${full_desktop_expected/\"segments\":\"project-assignment-notification,split-workbench,priority-filter,board-focus-view\"/\"segments\":\"project-assignment-notification,split-workbench,priority-filter,project-event-sources,board-focus-view\"}"
 
 assert_desktop_case "runner-only changes retain full coverage" \
@@ -737,6 +737,22 @@ wework_desktop_other_e2e=false
 wework_desktop_other_e2e_matrix={"include":[]}' \
   "wework/e2e/desktop/scenarios/collaboration-agent-automation-chain.scenario.mjs"
 
+assert_desktop_case "collaboration first-use scenario selects its core shard" \
+  'wework_desktop_e2e=true
+wework_desktop_core_e2e=true
+wework_desktop_core_e2e_matrix={"include":[{"id":"core-9","name":"Core / shard 9","segments":"collaboration-first-use"}]}
+wework_desktop_other_e2e=false
+wework_desktop_other_e2e_matrix={"include":[]}' \
+  "wework/e2e/desktop/scenarios/collaboration-first-use.scenario.mjs"
+
+assert_desktop_case "collaboration local agent capabilities scenario selects its core shard" \
+  'wework_desktop_e2e=true
+wework_desktop_core_e2e=true
+wework_desktop_core_e2e_matrix={"include":[{"id":"core-9","name":"Core / shard 9","segments":"collaboration-local-agent-capabilities"}]}
+wework_desktop_other_e2e=false
+wework_desktop_other_e2e_matrix={"include":[]}' \
+  "wework/e2e/desktop/scenarios/collaboration-local-agent-capabilities.scenario.mjs"
+
 assert_desktop_case "collaboration settings matrix scenario selects its core shard" \
   'wework_desktop_e2e=true
 wework_desktop_core_e2e=true

.github/workflows/e2e-tests.yml

@@ -431,6 +431,7 @@ jobs:
           ENVIRONMENT: development
           DB_AUTO_MIGRATE: "false"
           INIT_DATA_ENABLED: "true"
+          PLUGIN_LEGACY_UPLOAD_ENABLED: "true"
           RAG_RUNTIME_MODE: remote
           KNOWLEDGE_RUNTIME_URL: http://localhost:8200
           CHUNK_STORAGE_ENABLED: "true"
@@ -654,6 +655,20 @@ jobs:
           --health-timeout=5s
           --health-retries=3
 
+      minio:
+        image: bitnamilegacy/minio:2025.7.23-debian-12-r0
+        ports:
+          - 9000:9000
+        env:
+          MINIO_ROOT_USER: wegent-e2e
+          MINIO_ROOT_PASSWORD: wegent-e2e-secret
+        options: >-
+          --health-cmd="curl -f http://localhost:9000/minio/health/live"
+          --health-interval=10s
+          --health-timeout=5s
+          --health-start-period=20s
+          --health-retries=6
+
     steps:
       - name: Generate internal service token
         run: |
@@ -717,6 +732,17 @@ jobs:
       - name: Set up Node workspace
         uses: ./.github/actions/setup-node-workspace
 
+      - name: Restore bundled Codex cache
+        id: bundled-codex-cache
+        uses: actions/cache/restore@0057852bfaa89a56745cba8c7296529d2fc39830
+        with:
+          path: ~/.cache/wegent/codex
+          key: ${{ runner.os }}-wework-codex-v1-${{ hashFiles('wework/codex-binaries.lock.json') }}
+
+      - name: Prepare bundled Codex for local Executor
+        working-directory: ./wework
+        run: node scripts/prepare-codex-binary.mjs --materialize
+
       - name: Add local harness CLIs to PATH
         run: echo "$GITHUB_WORKSPACE/.github/claude-code-cli/node_modules/.bin" >> "$GITHUB_PATH"
 
@@ -835,6 +861,11 @@ jobs:
           ENVIRONMENT: development
           DB_AUTO_MIGRATE: "false"
           INIT_DATA_ENABLED: "true"
+          PLUGIN_LEGACY_UPLOAD_ENABLED: "true"
+          ATTACHMENT_S3_ENDPOINT: localhost:9000
+          ATTACHMENT_S3_ACCESS_KEY: wegent-e2e
+          ATTACHMENT_S3_SECRET_KEY: wegent-e2e-secret
+          ATTACHMENT_S3_USE_SSL: "false"
         run: |
           source .venv/bin/activate
           nohup uvicorn app.main:app --host 0.0.0.0 --port 8000 > backend.log 2>&1 &
@@ -867,7 +898,7 @@ jobs:
           pnpm exec tsx e2e/utils/mock-model-server.ts > mock-model-server.log 2>&1 &
           echo $! > mock-server.pid
 
-      - name: Start Local ClaudeCode Executor
+      - name: Start Local Executor
         env:
           EXECUTOR_MODE: local
           WEGENT_BACKEND_URL: http://localhost:8000
@@ -880,6 +911,9 @@ jobs:
           WORKSPACE_ROOT: ${{ runner.temp }}/wegent-workspace
           LOCAL_HEARTBEAT_INTERVAL: "5"
           LOG_LEVEL: info
+          CODEX_BINARY_PATH: ${{ github.workspace }}/wework/resources/binaries/codex/x86_64-unknown-linux-gnu/vendor/x86_64-unknown-linux-musl/bin/codex
+          CODEX_HOME: ${{ runner.temp }}/wegent-executor/codex
+          WEGENT_CODEX_HOME: ${{ runner.temp }}/wegent-executor/codex
           CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC: "1"
           CLAUDE_CODE_DISABLE_FEEDBACK_SURVEY: "1"
           CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK: "1"
@@ -940,6 +974,7 @@ jobs:
           PY
           )
           test -x executor/target/release/wegent-executor
+          test -x "$CODEX_BINARY_PATH"
           nohup env \
             -u GIT_TOKEN_AES_KEY \
             -u GIT_TOKEN_AES_IV \

backend/app/schemas/bot.py

@@ -19,6 +19,8 @@ class BotCreate(BaseModel):
     agent_config: dict[str, Any]
     system_prompt: Optional[str] = None
     mcp_servers: Optional[dict[str, Any]] = None
+    plugins: Optional[List[dict[str, Any]]] = None
+    capability_mode: Optional[str] = None
     default_knowledge_base_refs: Optional[List[KnowledgeBaseDefaultRef]] = None
     skills: Optional[List[str]] = None
     skill_refs: Optional[Dict[str, SkillRefMeta]] = None
@@ -43,6 +45,8 @@ class BotUpdate(BaseModel):
     agent_config: Optional[dict[str, Any]] = None
     system_prompt: Optional[str] = None
     mcp_servers: Optional[dict[str, Any]] = None
+    plugins: Optional[List[dict[str, Any]]] = None
+    capability_mode: Optional[str] = None
     default_knowledge_base_refs: Optional[List[KnowledgeBaseDefaultRef]] = None
     skills: Optional[List[str]] = None
     skill_refs: Optional[Dict[str, SkillRefMeta]] = None
@@ -69,6 +73,8 @@ class BotInDB(BaseModel):
     agent_config: dict[str, Any]
     system_prompt: Optional[str] = None
     mcp_servers: Optional[dict[str, Any]] = None
+    plugins: Optional[List[dict[str, Any]]] = None
+    capability_mode: str = "follow_device"
     default_knowledge_base_refs: Optional[List[KnowledgeBaseDefaultRef]] = None
     skills: Optional[List[str]] = None
     skill_refs: Optional[Dict[str, SkillRefMeta]] = None
@@ -97,6 +103,8 @@ class BotDetail(BaseModel):
     agent_config: dict[str, Any]
     system_prompt: Optional[str] = None
     mcp_servers: Optional[dict[str, Any]] = None
+    plugins: Optional[List[dict[str, Any]]] = None
+    capability_mode: str = "follow_device"
     default_knowledge_base_refs: Optional[List[KnowledgeBaseDefaultRef]] = None
     skills: Optional[List[str]] = None
     skill_refs: Optional[Dict[str, SkillRefMeta]] = None

backend/app/schemas/kind.py

@@ -181,6 +181,7 @@ class GhostSpec(BaseModel):
 
     systemPrompt: str
     mcpServers: Optional[Dict[str, Any]] = None
+    plugins: Optional[List[Dict[str, Any]]] = None
     defaultKnowledgeBaseRefs: Optional[List[KnowledgeBaseDefaultRef]] = None
     skills: Optional[List[str]] = None  # Skill names list
     preload_skills: Optional[List[str]] = Field(
@@ -462,6 +463,7 @@ class BotSpec(BaseModel):
     ghostRef: GhostRef
     shellRef: ShellRef
     modelRef: Optional[ModelRef] = None
+    capability_mode: str = "follow_device"
     secondaryModelRef: Optional[ModelRef] = Field(
         None,
         description="Secondary LLM model for auxiliary tasks. "

backend/app/services/adapters/bot_kinds.py

@@ -64,6 +64,20 @@ class BotKindsService(BaseService[Kind, BotCreate, BotUpdate]):
         # Add more sensitive keys here as needed
     ]
 
+    @staticmethod
+    def _resolve_create_capability_mode(obj_in: BotCreate) -> str:
+        if obj_in.capability_mode is not None:
+            return obj_in.capability_mode
+        has_explicit_capabilities = any(
+            (
+                obj_in.mcp_servers,
+                obj_in.plugins,
+                obj_in.skills,
+                obj_in.preload_skills,
+            )
+        )
+        return "manual" if has_explicit_capabilities else "follow_device"
+
     def _require_bot_permission(
         self,
         db: Session,
@@ -428,6 +442,7 @@ def create_with_user(
         ghost_spec = {
             "systemPrompt": obj_in.system_prompt or "",
             "mcpServers": obj_in.mcp_servers or {},
+            "plugins": obj_in.plugins or [],
         }
         if obj_in.default_knowledge_base_refs is not None:
             ghost_spec["defaultKnowledgeBaseRefs"] = [
@@ -555,6 +570,7 @@ def create_with_user(
             "ghostRef": {"name": ghost_name, "namespace": namespace},
             "shellRef": {"name": shell_ref_name, "namespace": shell_ref_namespace},
             "modelRef": {"name": model_ref_name, "namespace": model_ref_namespace},
+            "capability_mode": self._resolve_create_capability_mode(obj_in),
         }
         if obj_in.secondary_model_name:
             bot_spec["secondaryModelRef"] = {
@@ -1123,6 +1139,12 @@ def update_with_user(
             bot.json = bot_crd.model_dump()
             flag_modified(bot, "json")
 
+        if "capability_mode" in update_data:
+            bot_crd = Bot.model_validate(bot.json)
+            bot_crd.spec.capability_mode = update_data["capability_mode"]
+            bot.json = bot_crd.model_dump()
+            flag_modified(bot, "json")
+
         if "system_prompt" in update_data and ghost:
             ghost_crd = Ghost.model_validate(ghost.json)
             ghost_crd.spec.systemPrompt = update_data["system_prompt"] or ""
@@ -1136,6 +1158,13 @@ def update_with_user(
             flag_modified(ghost, "json")  # Mark JSON field as modified
             db.add(ghost)  # Add to session
 
+        if "plugins" in update_data and ghost:
+            ghost_crd = Ghost.model_validate(ghost.json)
+            ghost_crd.spec.plugins = update_data["plugins"] or []
+            ghost.json = ghost_crd.model_dump()
+            flag_modified(ghost, "json")
+            db.add(ghost)
+
         if "default_knowledge_base_refs" in update_data and ghost:
             ghost_crd = Ghost.model_validate(ghost.json)
             ghost_crd.spec.defaultKnowledgeBaseRefs = (
@@ -1697,6 +1726,7 @@ def _convert_to_bot_dict(
         # Extract data from components
         system_prompt = ""
         mcp_servers = {}
+        plugins = []
         shell_type = ""
         shell_name = ""
         agent_config = {}
@@ -1709,6 +1739,7 @@ def _convert_to_bot_dict(
             ghost_crd = Ghost.model_validate(ghost.json)
             system_prompt = ghost_crd.spec.systemPrompt
             mcp_servers = ghost_crd.spec.mcpServers or {}
+            plugins = ghost_crd.spec.plugins or []
 
         if shell and shell.json:
             shell_crd = Shell.model_validate(shell.json)
@@ -1861,6 +1892,8 @@ def _convert_to_bot_dict(
             "agent_config": agent_config,
             "system_prompt": system_prompt,
             "mcp_servers": mcp_servers,
+            "plugins": plugins,
+            "capability_mode": bot_crd.spec.capability_mode,
             "default_knowledge_base_refs": default_knowledge_base_refs,
             "skills": skills,
             "skill_refs": skill_refs,
@@ -2428,6 +2461,8 @@ def clone_bot(
             skill_refs=skill_refs_meta,
             preload_skills=bot_dict.get("preload_skills"),
             preload_skill_refs=preload_skill_refs_meta,
+            plugins=bot_dict.get("plugins"),
+            capability_mode=bot_dict.get("capability_mode", "follow_device"),
             namespace=namespace,
         )