AI agents invoke host_exec to trigger actions in Ns Hpc. 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.
host_exec executes commands on the host system, making it Execute category. Severity is high because even 'pre-configured' commands can have broad effects in an HPC environment (resource consumption, data access, system state changes), though the sandboxing and pre-configuration constraints mitigate it from critical.
From the tool's definition Tool description states it can 'Run a pre-configured host command', which executes external operations. The server description emphasizes 'sandboxed interface' and 'bubblewrap isolation', indicating awareness that execution capabilities pose risks requiring…
Attacks that exploit this kind of access
Run a pre-configured host command or list available commands. It is categorised as a Execute tool in the Ns Hpc MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Ns Hpc MCP server in PolicyLayer and add a rule for host_exec: 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 Ns Hpc. Nothing to install.
host_exec 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 host_exec 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 host_exec. 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.
host_exec is provided by the Ns Hpc MCP server (li-yq/namespaced-hpc-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Every MCP server has a record like this.
Type a name, get the same breakdown: verified identity, auth posture, risk grade, capabilities, recommended policy.
Teams ship this data inside their own products. See what a licence covers →