sandbox_search
BM25-ranked full-text search across all sandboxed content. Pass multiple queries as an array to batch all questions in one call. Use after sandbox_ingest or sandbox_execute to retrieve relevant snippets without pulling raw data into context.
This record as markdown: /tools/io-github-homenshum-nodebench/sandbox-search.md
What sandbox_search does on Nodebench
AI agents call sandbox_search to retrieve information from Nodebench without modifying anything. It is typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
Why sandbox_search is rated Low
sandbox_search is a search/query operation that retrieves data from sandboxed content without side effects. It matches the Read category pattern of querying and fetching information. The low severity reflects that search operations have minimal blast radius—an AI agent misusing this would at worst retrieve unintended information, not modify or destroy data.
From the tool's definition Tool performs 'full-text search' and 'retrieve relevant snippets' with no modification, deletion, or execution capabilities described. The explicit purpose is information retrieval after other operations have been performed.
Attacks that exploit this kind of access
The rule that runs sandbox_search safely
PolicyLayer is an MCP gateway: it sits between your AI agents and Nodebench, and checks every tool call against a rule you set before the call runs. Nothing changes on the server itself. For sandbox_search, this is the rule to start with:
sandbox_search is read-only, so it stays allowed. Everything else on the server is denied unless you say otherwise.
The button opens the PolicyLayer dashboard: create your workspace, connect Nodebench, apply this rule, and every sandbox_search call is checked against it from then on.
Questions about sandbox_search
BM25-ranked full-text search across all sandboxed content. Pass multiple queries as an array to batch all questions in one call. Use after sandbox_ingest or sandbox_execute to retrieve relevant snippets without pulling raw data into context. It is categorised as a Read tool in the Nodebench MCP Server, which means it retrieves data without modifying state.
Register the Nodebench MCP server in PolicyLayer and add a rule for sandbox_search: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Nodebench. Nothing to install.
sandbox_search is a Read tool with low risk. Read-only tools are generally safe to allow by default.
Yes. Add a rate_limit block to the sandbox_search rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for sandbox_search. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
sandbox_search is provided by the Nodebench MCP server (nodebench-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
More on Nodebench, and thousands of servers like it.
This server
Across the catalogue