Show HN: Stroq – block commands injected through MCP tool output (stroq.dev)

Thanks! Stroq connects to your code-writing agent via tools. It monitors the data the agent reads and, before a command or MCP call is executed, checks whether the action was triggered by untrusted content. Stroq can then request confirmation or block the action, as well as explain the reason.

everything runs locally

you can try it with @stroq/cli init

let me know which coding agent do you use :)