Skip to content

Standardize on a sandbox execution substrate for agent runs #56

Description

@stubbi

Note: execution substrate for agent runs

crabfleet is mission control for agent runs. A recurring question across the OpenClaw agent stack is where those runs actually execute and how they are isolated and torn down.

Worth considering whether crabfleet should standardize on a sandbox execution layer rather than leaving it per agent:

This is a lightweight pointer, not a worked proposal. Filed alongside related issues on mitos, crabbox, paperclip, and clownfish to keep the layering visible:

mitos      snapshot-fork microVM engine
crabbox    lease a box, sync, run
crabfleet  mission control for agent runs

If crabfleet already delegates execution to crabbox or similar, close this as already covered.

Metadata

Metadata

Assignees

No one assigned

    Labels

    P3Low-risk cleanup, docs, polish, ergonomics, or speculative feature.impact:securityThis issue is about security boundaries, credentials, authz, sandboxing, or sensitive data.

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions