sandbox_execute
Run a shell command, automatically index the output into the sandbox, and return only a summary. The raw stdout/stderr stays in SQLite — only line counts and a preview enter context. Use instead of raw shell execution for commands that produce large output (build logs, test results, file listings...
This record as markdown: /tools/io-github-homenshum-nodebench/sandbox-execute.md
What sandbox_execute does on Nodebench
AI agents invoke sandbox_execute to trigger actions in Nodebench. What it does depends on the arguments the agent supplies, and its effects often reach beyond the immediate call: builds kicked off, notifications sent, workflows started.
Why sandbox_execute is rated High
This tool executes shell commands, which can have broad and unpredictable effects depending on the arguments provided. While the tool attempts to mitigate information leakage by storing raw output separately and returning only summaries, the ability to run arbitrary shell commands on the sandboxed system is a high-severity Execute action.
From the tool's definition Tool explicitly states it will 'Run a shell command' and 'execute' operations. The description confirms it executes arbitrary shell commands with automatic indexing of output.
Attacks that exploit this kind of access
The rule that runs sandbox_execute 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_execute, this is the rule to start with:
sandbox_execute stays usable, but rate-capped: a runaway agent can't fire it dozens of times a minute. 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_execute call is checked against it from then on.
Questions about sandbox_execute
Run a shell command, automatically index the output into the sandbox, and return only a summary. The raw stdout/stderr stays in SQLite — only line counts and a preview enter context. Use instead of raw shell execution for commands that produce large output (build logs, test results, file listings, git logs). It is categorised as a Execute tool in the Nodebench MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Nodebench MCP server in PolicyLayer and add a rule for sandbox_execute: 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_execute is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.
Yes. Add a rate_limit block to the sandbox_execute 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_execute. 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_execute 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.
Across the catalogue