Pre-admission
Inspect third-party plugin metadata, dependencies, requested filesystem/network access, known indicators and suspicious installation steps.
Read the evidence and analysis →Tomorrow’s agent governance will need plugin admission decisions, runtime telemetry and accountable incident workflows. Here is the roadmap, verification plan, and its limits.
Proposed system, not product claim
We intend to develop specialized Security Officers that can inspect plugin packages before an agent is allowed to use them, review signatures and provenance, flag unexpected privileges, and correlate behavior with requests made to rootless enforcement services. These are planned capabilities, not currently shipped protections.
Inspect third-party plugin metadata, dependencies, requested filesystem/network access, known indicators and suspicious installation steps.
Read the evidence and analysis →Collect permitted service, container, process and network events; correlate to action request IDs without placing agents in unrestricted host authority.
Read the evidence and analysis →Create a bounded incident report and propose deny, quarantine or approval actions. Automatic intervention is a policy decision requiring testing.
Read the evidence and analysis →ROADMAP / NOT SHIPPED
The potential Security Officer-specific rootless daemon is a separate, future feature. It is not the present XMAN daemon, and it does not currently provide runtime interception, automatic kill-switch behavior, plugin malware classification or an anti-exfiltration guarantee. A design might include a narrowly privileged event subscriber, policy-checking API, rootless process inspection where the kernel permits it, and measured escalation to the operator. The effective boundary depends on the actual privileges, namespaces, mount and network configuration used.
Any future integration with OpenShell or another enforcement service must be explicitly tested for privilege separation and false-positive recovery. We do not claim any particular certification.
Future architecture · separate component · alpha disclosure
A future Security Officer may communicate with its own independently deployed rootless monitoring and enforcement daemon. This component would be separate from today's XMAN manager daemon, with its own limited Unix identity, narrowly scoped event sources, and explicitly granted actions. The agent can flag and recommend; deterministic policy and an authorized operator must decide whether a quarantine, deny rule, or other intervention is permitted.
Proposed plugin provenance, signatures, dependency scanning, policy review and quarantine suggestions. No claim of universal malicious-code detection.
Collect permitted container, process, filesystem, network and approval signals through a least-privilege, rootless service; monitor blind spots explicitly.
Produce timestamped observations, linked request IDs, reproducible incident records and bounded recommendations. Automatic action needs its own policy and tests.
We respect the spirit of OSSTMM—transparent methodology, defined scope, controlled testing and measurable evidence. We do not claim OSSTMM certification, formal conformity, or any other third-party security certification for Clawsolator at this time.
Threat model, environment, configuration, expected enforcement point, and approval identity.
Host logs, daemon requests, sandbox policy decisions and network traces under controlled conditions.
Pass/fail, affected versions, edge cases, remediations, and how anyone can reproduce the test.
We will not claim an AI can recognize every malicious skill, catch all prompt injections, or make a deployment “impeccable.” False positives, blind spots, OS vulnerabilities, credentials, allowed egress, and policy errors still matter. Security is tested engineering, not omniscience.
Explore the evidence
Inspect the controls, verify the boundaries, and ask what happens after a failure—not merely before it.
Explore the security lattice →Read the architecture →CANONICAL TRUST BOUNDARY
Each registered agent uses the authenticated requester API through the trusted Gateway bridge. The XMAN management daemon governs policy, resource admission and durable outcomes through its private Unix-socket interface. The operator API is separate; the master coordinates but does not become the operator. The future SysOp and Security Officer daemons would be separate, permissioned services, not current Alpha functionality.