Spectra Assure Free Trial
Get your 14-day free trial of Spectra Assure for Software Supply Chain Security
Get Free TrialMore about Spectra Assure Free Trial
NVIDIA governs what agents do. ReversingLabs verifies what agents use.
Autonomous agents download packages, install skills, load models, execute software, and move files as part of normal operation. Each artifact becomes part of the agent's effective attack surface.
NVIDIA OpenShell provides sandboxed execution and policy controls that govern what an agent is permitted to do and what resources it can access.
ReversingLabs adds intelligence about the software, packages, models, and files those agents interact with, giving security teams evidence about the artifacts entering and moving through agent workflows.
Together, policy and artifact intelligence can help organizations make more informed decisions before software is executed, while agents are operating, and when security teams investigate activity.
Inspect What Agents Use
Analyze files, packages, models, and other artifacts before they’re trusted.
How it Works
Admission control · Day one and pilot
Admit only what has been analyzed
Spectra Assure analyzes every sandbox image and SKILL.md bundle at the binary level. Only approved digests reach the registry OpenShell builds sandboxes from, and a governance interceptor can make sure policy can't be quietly weakened later (pilot).
Pull only what has passed analysis
Packages and models come from your internal repository, gated by Spectra Assure and the Catalog add-on. Components that fail analysis never reach the sandbox, and the findings go to the team that requested them.
Runtime inspection · Pilot
Inspect what agents send and receive
File Inspection Engine runs as OpenShell supervisor middleware. It checks outbound bodies before credentials are injected and, with OpenShell 0.1.x, downloads up to 4 MiB before the agent sees them. It is designed to resolve most requests by hash and file type in milliseconds.
Review what agents ask for
NVIDIA leaves the approval of new access to the organization. When a blocked agent proposes new network access, an RL reviewer checks the destination's reputation before the rule is approved, closing the gap NVIDIA documents for automatic approval of new public hosts.
Why ReversingLabs
40B+ File Intelligence
Tap into one of the world's largest file reputation corpora to add context, history, and known-good / known-bad intelligence to every inspection.
Trust, but decompile.
NVIDIA built the runtime that decides what agents can do. Give it the evidence
to decide well. Talk to us about a pilot in your agent environment.
