Makes packages available in the environment and optionally runs a command. Returns stdout, stderr, exit code, and duration.
AI agents invoke shell to trigger actions in Http. 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.
The tool explicitly runs commands and returns stdout/stderr/exit code, which is the hallmark of arbitrary shell execution. An AI agent could use this to run any command on the host system, making it critical severity. The 'makes packages available' aspect also implies environment modification (Write), but arbitrary command execution is the dominant and more severe capability.
From the tool's definition 'optionally runs a command. Returns stdout, stderr, exit code, and duration' — this tool executes arbitrary shell commands in the environment
Documented attack patterns abuse exactly the kind of access shell gives an agent:
PolicyLayer is an MCP gateway — it sits between your AI agents and Http, and nothing reaches the server without passing your rules. This is the rule we recommend for shell:
{
"version": "1",
"default": "deny",
"tools": {
"shell": {
"limits": [
{
"counter": "shell_rate",
"window": "minute",
"max": 10,
"scope": "grant"
}
]
}
}
}shell 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.
Free to start. No card required.
Makes packages available in the environment and optionally runs a command. Returns stdout, stderr, exit code, and duration. It is categorised as a Execute tool in the Http MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Http MCP server in PolicyLayer and add a rule for shell: 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 Http. Nothing to install.
shell 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 shell 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 shell. 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.
shell is provided by the Http MCP server (@paretools/http). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Start from Http, add the rest of your stack, and see everything your agents can call. Then put policy on all of it.
Free to start. No card required.
202 Http tools catalogued and risk-classified — across an index of 43,000+ MCP servers.