Skip to main content

2 posts tagged with "sandboxing"

View All Tags

The Part of Your App You Can Hand to Something You Don't Trust

CJ Cummings
Creator and maintainer of Stof, co-founder & CEO of Limitr

Sometime in the last year, "the agent will just edit the file" stopped being hypothetical and became a normal workflow. Point an agent at a config, a document, or a blob of state, and it reads, edits, and writes it back, with no UI in between.

That's useful. It's also a trust problem most formats were never built to handle.

How to Safely Give AI Agents New Tools at Runtime

CJ Cummings
Creator and maintainer of Stof, co-founder & CEO of Limitr

Most systems treat "let an agent gain new tools while it runs" and "stop an agent from doing damage" as two problems with two solutions: a tool registry plus an approval workflow, or a plugin marketplace plus a permission system added afterward.

In Stof, they're the same mechanism. A document can only reach what it has been given, so a tool that arrives at runtime is limited the same way everything else in the document already is. Not because a permission check ran, but because there's nothing else there to reach.