fix(sandbox): use updated WorkspaceSpec when resuming from persisted state

[Buktal]

AgentScope-Java Version

2.0.0-SNAPSHOT

Description

Background:
When an application modifies WORKSPACE_PROJECTION_ROOTS in SandboxFilesystemSpec (e.g., adding .skills-cache to the projection list), the change is ignored for existing sandboxes resumed from Redis. The sandbox continues using the old WorkspaceSpec persisted in SandboxState.

Root cause:
SandboxManager.acquire() deserializes SandboxState from Redis and directly passes it to client.resume(state). The WorkspaceSpec inside state was captured during the first create() call and is never updated.

Changes made:

• Modified SandboxManager.acquire() (L106-109) to overwrite state.workspaceSpec with the current sandboxContext.workspaceSpec before resuming
• This ensures application-layer configuration changes take effect without requiring manual cache deletion

How to test:

1. Deploy an application with initial projection roots: ["A", "B", "C"]
2. Run the application, create a sandbox (state persisted to Redis)
3. Update projection roots to ["A", "B", "C", "D"] where D exists on host
4. Redeploy and verify that directory D appears in the sandbox without manual cache deletion

Related:

• Fixes [Bug]: Sandbox resume ignores updated WorkspaceSpec configuration #1926

Checklist

• Code has been formatted with mvn spotless:apply
• All tests are passing (mvn test)
• Javadoc comments are complete and follow project conventions
• Related documentation has been updated (e.g. links, examples, etc.)
• Code is ready for review

[itxaiohanglover]

Nice fix — the null check on workspaceSpec and the defensive .copy() are solid. One thing I noticed: it'd be great to have a unit test covering the resume path to make sure the stale spec actually gets overwritten. Other than that, looks good to me.