Skip to content
Open
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 8 additions & 3 deletions content/manuals/ai/sandboxes/security/isolation.md
Original file line number Diff line number Diff line change
Expand Up @@ -148,6 +148,13 @@ never works directly against your host repository. Even with full root
inside the VM, it cannot modify your `.git` directory, your working tree,
or any tracked file on your host.

> [!IMPORTANT]
> Clone mode protects your host repository from modification, **not from
> inspection**. Your repository is still mounted read-only into the sandbox,
> including untracked files and files excluded by `.gitignore`. Files such as
> `.env` remain readable by the agent. Store secrets outside your working
> directory or use [credential isolation](credentials.md) instead.

```mermaid
flowchart LR
subgraph host["Host repository (untouched)"]
Expand All @@ -174,9 +181,7 @@ How the boundary is enforced:
untracked files and files excluded by `.gitignore`. Nothing the agent
does inside the VM can write back through that mount, but all files
under the Git root are readable inside the sandbox. This includes
credential files not tracked by Git, such as `.env`. Store
secrets outside your working directory or use
[credential isolation](credentials.md) instead.
credential files not tracked by Git, such as `.env`.
- The agent works on a private clone that lives inside the sandbox. The
clone has its own index, its own refs, and its own working tree. Writes
to the clone never reach your host.
Expand Down