Hermetic 'blinded' agents using Starlark sandbox
Starlark sandbox + opaque values defends agents from prompt injection, not just wrapping smolagents.
Portable, self-contained sandboxes for AI coding agents - microVMs, containers, isolation harnesses.
Useful curated list of agent sandbox options, but it's just links with no implementation.
Developers building or deploying AI coding agents
Starlark sandbox + opaque values defends agents from prompt injection, not just wrapping smolagents.
Open-source alternative to Ramp Inspect with sandboxed agent execution and review workflows.
70+ skills, but it's a GitHub directory — no execution engine or platform.
OCI-backed .agent archives beat zip files for sharing agent setups.
202 tips from 10 experts, but it's just a curated list.
AST-validated shell commands with OS sandboxing for safer AI coding.