Merge pull request #5028 from w31r4/feat/neo-skill-self-iteration

feat: 接入 Shipyard Neo 自迭代 Skill 闭环与管理能力

Author: RC-CHN

.gitignore

@@ -54,3 +54,7 @@ IFLOW.md
 # genie_tts data
 CharacterModels/
 GenieData/
+.agent/
+.codex/
+.opencode/
+.kilocode/

CONTRIBUTING.md

@@ -46,6 +46,32 @@ ruff check .
 
 如果您使用 VSCode，可以安装 `Ruff` 插件。
 
+##### PR 功能完整性验证（推荐）
+
+如果您希望在本地做一套接近 CI 的完整验证，可使用：
+
+```bash
+make pr-test-neo
+```
+
+该命令会执行：
+- `uv sync --group dev`
+- `ruff format --check .` 与 `ruff check .`
+- Neo 相关关键测试
+- `main.py` 启动 smoke test（检测 `http://localhost:6185`）
+
+需要全量验证时可使用：
+
+```bash
+make pr-test-full
+```
+
+如果只想快速重复执行（跳过依赖同步和 dashboard 构建）：
+
+```bash
+make pr-test-full-fast
+```
+
 
 ## Contributing Guide
 
@@ -88,3 +114,29 @@ We use Ruff as our code formatter and static analysis tool. Before submitting yo
 ruff format .
 ruff check .
 ```
+
+##### PR completeness checks (recommended)
+
+To run a local validation flow close to CI, use:
+
+```bash
+make pr-test-neo
+```
+
+This command runs:
+- `uv sync --group dev`
+- `ruff format --check .` and `ruff check .`
+- Neo-related critical tests
+- a startup smoke test against `http://localhost:6185`
+
+For full validation, use:
+
+```bash
+make pr-test-full
+```
+
+For faster repeated runs (skip dependency sync and dashboard build), use:
+
+```bash
+make pr-test-full-fast
+```

Makefile

@@ -1,4 +1,4 @@
-.PHONY: worktree worktree-add worktree-rm
+.PHONY: worktree worktree-add worktree-rm pr-test-neo pr-test-full pr-test-full-fast
 
 WORKTREE_DIR ?= ../astrbot_worktree
 BRANCH ?= $(word 2,$(MAKECMDGOALS))
@@ -27,6 +27,15 @@ endif
 		echo "Worktree $(WORKTREE_DIR)/$(BRANCH) not found."; \
 	fi
 
+pr-test-neo:
+	./scripts/pr_test_env.sh --profile neo
+
+pr-test-full:
+	./scripts/pr_test_env.sh --profile full
+
+pr-test-full-fast:
+	./scripts/pr_test_env.sh --profile full --skip-sync --no-dashboard
+
 # Swallow extra args (branch/base) so make doesn't treat them as targets
 %:
 	@true

astrbot/core/astr_main_agent.py

@@ -20,18 +20,32 @@
 from astrbot.core.astr_agent_run_util import AgentRunner
 from astrbot.core.astr_agent_tool_exec import FunctionToolExecutor
 from astrbot.core.astr_main_agent_resources import (
+    ANNOTATE_EXECUTION_TOOL,
+    BROWSER_BATCH_EXEC_TOOL,
+    BROWSER_EXEC_TOOL,
     CHATUI_SPECIAL_DEFAULT_PERSONA_PROMPT,
+    CREATE_SKILL_CANDIDATE_TOOL,
+    CREATE_SKILL_PAYLOAD_TOOL,
+    EVALUATE_SKILL_CANDIDATE_TOOL,
     EXECUTE_SHELL_TOOL,
     FILE_DOWNLOAD_TOOL,
     FILE_UPLOAD_TOOL,
+    GET_EXECUTION_HISTORY_TOOL,
+    GET_SKILL_PAYLOAD_TOOL,
     KNOWLEDGE_BASE_QUERY_TOOL,
+    LIST_SKILL_CANDIDATES_TOOL,
+    LIST_SKILL_RELEASES_TOOL,
     LIVE_MODE_SYSTEM_PROMPT,
     LLM_SAFETY_MODE_SYSTEM_PROMPT,
     LOCAL_EXECUTE_SHELL_TOOL,
     LOCAL_PYTHON_TOOL,
+    PROMOTE_SKILL_CANDIDATE_TOOL,
     PYTHON_TOOL,
+    ROLLBACK_SKILL_RELEASE_TOOL,
+    RUN_BROWSER_SKILL_TOOL,
     SANDBOX_MODE_PROMPT,
     SEND_MESSAGE_TO_USER_TOOL,
+    SYNC_SKILL_RELEASE_TOOL,
     TOOL_CALL_PROMPT,
     TOOL_CALL_PROMPT_SKILLS_LIKE_MODE,
     retrieve_knowledge_base,
@@ -832,19 +846,73 @@ def _apply_sandbox_tools(
 ) -> None:
     if req.func_tool is None:
         req.func_tool = ToolSet()
-    if config.sandbox_cfg.get("booter") == "shipyard":
+    booter = config.sandbox_cfg.get("booter", "shipyard_neo")
+    if booter == "shipyard":
         ep = config.sandbox_cfg.get("shipyard_endpoint", "")
         at = config.sandbox_cfg.get("shipyard_access_token", "")
         if not ep or not at:
             logger.error("Shipyard sandbox configuration is incomplete.")
             return
         os.environ["SHIPYARD_ENDPOINT"] = ep
         os.environ["SHIPYARD_ACCESS_TOKEN"] = at
+
     req.func_tool.add_tool(EXECUTE_SHELL_TOOL)
     req.func_tool.add_tool(PYTHON_TOOL)
     req.func_tool.add_tool(FILE_UPLOAD_TOOL)
     req.func_tool.add_tool(FILE_DOWNLOAD_TOOL)
-    req.system_prompt = f"{req.system_prompt}\n{SANDBOX_MODE_PROMPT}\n"
+    if booter == "shipyard_neo":
+        # Neo-specific path rule: filesystem tools operate relative to sandbox
+        # workspace root. Do not prepend "/workspace".
+        req.system_prompt += (
+            "\n[Shipyard Neo File Path Rule]\n"
+            "When using sandbox filesystem tools (upload/download/read/write/list/delete), "
+            "always pass paths relative to the sandbox workspace root. "
+            "Example: use `baidu_homepage.png` instead of `/workspace/baidu_homepage.png`.\n"
+        )
+
+        req.system_prompt += (
+            "\n[Neo Skill Lifecycle Workflow]\n"
+            "When user asks to create/update a reusable skill in Neo mode, use lifecycle tools instead of directly writing local skill folders.\n"
+            "Preferred sequence:\n"
+            "1) Use `astrbot_create_skill_payload` to store canonical payload content and get `payload_ref`.\n"
+            "2) Use `astrbot_create_skill_candidate` with `skill_key` + `source_execution_ids` (and optional `payload_ref`) to create a candidate.\n"
+            "3) Use `astrbot_promote_skill_candidate` to release: `stage=canary` for trial; `stage=stable` for production.\n"
+            "For stable release, set `sync_to_local=true` to sync `payload.skill_markdown` into local `SKILL.md`.\n"
+            "Do not treat ad-hoc generated files as reusable Neo skills unless they are captured via payload/candidate/release.\n"
+            "To update an existing skill, create a new payload/candidate and promote a new release version; avoid patching old local folders directly.\n"
+        )
+
+        # Determine sandbox capabilities from an already-booted session.
+        # If no session exists yet (first request), capabilities is None
+        # and we register all tools conservatively.
+        from astrbot.core.computer.computer_client import session_booter
+
+        sandbox_capabilities: list[str] | None = None
+        existing_booter = session_booter.get(session_id)
+        if existing_booter is not None:
+            sandbox_capabilities = getattr(existing_booter, "capabilities", None)
+
+        # Browser tools: only register if profile supports browser
+        # (or if capabilities are unknown because sandbox hasn't booted yet)
+        if sandbox_capabilities is None or "browser" in sandbox_capabilities:
+            req.func_tool.add_tool(BROWSER_EXEC_TOOL)
+            req.func_tool.add_tool(BROWSER_BATCH_EXEC_TOOL)
+            req.func_tool.add_tool(RUN_BROWSER_SKILL_TOOL)
+
+        # Neo-specific tools (always available for shipyard_neo)
+        req.func_tool.add_tool(GET_EXECUTION_HISTORY_TOOL)
+        req.func_tool.add_tool(ANNOTATE_EXECUTION_TOOL)
+        req.func_tool.add_tool(CREATE_SKILL_PAYLOAD_TOOL)
+        req.func_tool.add_tool(GET_SKILL_PAYLOAD_TOOL)
+        req.func_tool.add_tool(CREATE_SKILL_CANDIDATE_TOOL)
+        req.func_tool.add_tool(LIST_SKILL_CANDIDATES_TOOL)
+        req.func_tool.add_tool(EVALUATE_SKILL_CANDIDATE_TOOL)
+        req.func_tool.add_tool(PROMOTE_SKILL_CANDIDATE_TOOL)
+        req.func_tool.add_tool(LIST_SKILL_RELEASES_TOOL)
+        req.func_tool.add_tool(ROLLBACK_SKILL_RELEASE_TOOL)
+        req.func_tool.add_tool(SYNC_SKILL_RELEASE_TOOL)
+
+    req.system_prompt = f"{req.system_prompt or ''}\n{SANDBOX_MODE_PROMPT}\n"
 
 
 def _proactive_cron_job_tools(req: ProviderRequest) -> None:

astrbot/core/astr_main_agent_resources.py

@@ -13,11 +13,25 @@
 from astrbot.core.astr_agent_context import AstrAgentContext
 from astrbot.core.computer.computer_client import get_booter
 from astrbot.core.computer.tools import (
+    AnnotateExecutionTool,
+    BrowserBatchExecTool,
+    BrowserExecTool,
+    CreateSkillCandidateTool,
+    CreateSkillPayloadTool,
+    EvaluateSkillCandidateTool,
     ExecuteShellTool,
     FileDownloadTool,
     FileUploadTool,
+    GetExecutionHistoryTool,
+    GetSkillPayloadTool,
+    ListSkillCandidatesTool,
+    ListSkillReleasesTool,
     LocalPythonTool,
+    PromoteSkillCandidateTool,
     PythonTool,
+    RollbackSkillReleaseTool,
+    RunBrowserSkillTool,
+    SyncSkillReleaseTool,
 )
 from astrbot.core.message.message_event_result import MessageChain
 from astrbot.core.platform.message_session import MessageSession
@@ -449,6 +463,20 @@ async def retrieve_knowledge_base(
 LOCAL_PYTHON_TOOL = LocalPythonTool()
 FILE_UPLOAD_TOOL = FileUploadTool()
 FILE_DOWNLOAD_TOOL = FileDownloadTool()
+BROWSER_EXEC_TOOL = BrowserExecTool()
+BROWSER_BATCH_EXEC_TOOL = BrowserBatchExecTool()
+RUN_BROWSER_SKILL_TOOL = RunBrowserSkillTool()
+GET_EXECUTION_HISTORY_TOOL = GetExecutionHistoryTool()
+ANNOTATE_EXECUTION_TOOL = AnnotateExecutionTool()
+CREATE_SKILL_PAYLOAD_TOOL = CreateSkillPayloadTool()
+GET_SKILL_PAYLOAD_TOOL = GetSkillPayloadTool()
+CREATE_SKILL_CANDIDATE_TOOL = CreateSkillCandidateTool()
+LIST_SKILL_CANDIDATES_TOOL = ListSkillCandidatesTool()
+EVALUATE_SKILL_CANDIDATE_TOOL = EvaluateSkillCandidateTool()
+PROMOTE_SKILL_CANDIDATE_TOOL = PromoteSkillCandidateTool()
+LIST_SKILL_RELEASES_TOOL = ListSkillReleasesTool()
+ROLLBACK_SKILL_RELEASE_TOOL = RollbackSkillReleaseTool()
+SYNC_SKILL_RELEASE_TOOL = SyncSkillReleaseTool()
 
 # we prevent astrbot from connecting to known malicious hosts
 # these hosts are base64 encoded

astrbot/core/computer/booters/base.py

@@ -1,4 +1,9 @@
-from ..olayer import FileSystemComponent, PythonComponent, ShellComponent
+from ..olayer import (
+    BrowserComponent,
+    FileSystemComponent,
+    PythonComponent,
+    ShellComponent,
+)
 
 
 class ComputerBooter:
@@ -11,6 +16,19 @@ def python(self) -> PythonComponent: ...
     @property
     def shell(self) -> ShellComponent: ...
 
+    @property
+    def capabilities(self) -> tuple[str, ...] | None:
+        """Sandbox capabilities (e.g. ('python', 'shell', 'filesystem', 'browser')).
+
+        Returns None if the booter doesn't support capability introspection
+        (backward-compatible default).  Subclasses override after boot.
+        """
+        return None
+
+    @property
+    def browser(self) -> BrowserComponent | None:
+        return None
+
     async def boot(self, session_id: str) -> None: ...
 
     async def shutdown(self) -> None: ...