Prevent Path Traversal in Host Tools for AI Agents
An agent can write outside its job through a privileged file helper. Let trusted code choose destinations and authorize report downloads.
An agent can cause damage through an overly powerful host tool even when its VM stays isolated. These articles separate that mistake from a runtime escape and show where to enforce the permission, lifecycle, or containment rule.
An agent can write outside its job through a privileged file helper. Let trusted code choose destinations and authorize report downloads.
Reject a model-selected hostname before lookup, check every returned address, and preserve the chosen IP and TLS hostname in a Node HTTPS request.
Compare two Python launches using a dummy token and file descriptor, then restrict the environment and handles passed to the child.
Keep a filename out of the shell, then handle spaces and leading options with a fixed converter and an explicit argument list.
Give a trusted Rust file tool one directory handle. Check allowed reads, traversal, and outward symlinks without claiming the library confines arbitrary code.
Follow a control request from its credential to a job inspection. Localhost and Host headers cannot replace authentication or authorize another job.
Follow a constructed README attack from a proposed export to a denied tool call, and check that the downstream service performs no forbidden effect.
Compare waiting for one process with cleanup on completion, timeout, cancellation, and launcher failure. Give the whole job an independent supervisor.