feat: Initial library extraction from PIN LLM benchmark

DockerSandbox + LangChain file/shell tools extracted into a standalone package.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

docker_agent_sandbox/tools/list_dir.py

@@ -0,0 +1,32 @@
+"""list_dir.py – tool for listing directory contents inside the sandbox."""
+
+from __future__ import annotations
+
+from shlex import quote
+from typing import TYPE_CHECKING
+
+from langchain_core.tools import BaseTool, tool
+from loguru import logger
+
+if TYPE_CHECKING:
+    from docker_agent_sandbox.sandbox import DockerSandbox
+
+
+def make_list_dir_tool(sandbox: "DockerSandbox") -> BaseTool:
+    """Return a list_dir tool bound to *sandbox*."""
+
+    @tool
+    def list_dir(path: str = ".") -> str:
+        """
+        List the contents of a directory at *path*.
+
+        Returns ``ls -lA`` output,
+        or an error message if the path does not exist or is not a directory.
+        """
+        logger.debug("Listing files inside sandbox: {}", path)
+        exit_code, output = sandbox.exec(f"ls -lA -- {quote(path)}")
+        if exit_code != 0:
+            return f"[ERROR listing {path!r}] {output.strip()}"
+        return output
+
+    return list_dir