Scanned repository says: NVIDIA NemoClaw is an open source reference stack for running supported AI agents more safely inside NVIDIA OpenShell sandboxes.
Metal Mantra scanned this public repository on its own initiative; the maintainer did not submit it. Maintainers: scan the repository yourself to take this report over, or contact us.
Inspect GitHub sourceVotes show community interest, not safety. They never change this score.
Critical finding · details withheld for 30 days
Inspect the findingsRuntime safety, real-world agent capability, fitness for your use case or independent certification.
Structured, cited information from the scanned archive. Repository-authored claims are labeled and never treated as verified capabilities.
NVIDIA NemoClaw is an open source reference stack for running supported AI agents more safely inside NVIDIA OpenShell sandboxes.
README.md:12It provides guided onboarding, managed inference, network policy, managed integrations, OpenShell-backed state persistence, and lifecycle operations through the NemoClaw CLI and its agent-specific aliases.
README.md:13OpenClaw (default)
README.md:17No explicit capability or integration was extracted; that does not mean none exists.
Manifest declares Node.js >=22.19.0 as a requirement.
package.json:161Declares a dependency on @aws-sdk/client-bedrock-runtime.
package.json:112Declares a dependency on @oclif/core.
package.json:113Declares a dependency on ajv.
package.json:114Declares a dependency on execa.
package.json:115Declares a dependency on js-yaml.
package.json:117Declares a dependency on json5.
package.json:116NVIDIA takes security seriously.
README.md:103Manifest declares Apache-2.0 as its license.
package.json:6Submit a report through the NVIDIA Vulnerability Disclosure Program.
README.md:107Send an email to psirt@nvidia.com encrypted with the NVIDIA PGP key.
README.md:108These are references in selected source files, not proof that a feature works at runtime.
Read from commit ccf8e4e at Oct 12, 2026, 4:59 AM UTC. Source links open that exact commit. Repository statements are unreviewed and do not affect the score.
Critical details are withheld publicly for the first 30 days after a scan. A matched pattern is a lead for investigation, not a confirmed vulnerability.
Dependencies: 136 exact runtime versions from package-lock.json checked against OSV.dev on Oct 12, 2026; 2 with known advisories. Reachability is not checked.
Details are withheld from public view for 30 days after the scan. The owner can see them in full.
Review the private assessment and address the finding.
Details are withheld from public view for 30 days after the scan. The owner can see them in full.
Review the private assessment and address the finding.
The code loads data with a format that can run code while it is read, such as pickle or unrestricted YAML. A poisoned file, cache entry or downloaded model then runs with the agent's permissions.
Use a data-only format (JSON, safetensors, yaml.safe_load) for anything that is not fully trusted.
Metal Mantra could not confirm that the owner controls the stated domain. The domain may belong to someone else.
Verify the domain using the Metal Mantra DNS TXT challenge.
No privacy policy link was found on the stated homepage. Buyers cannot see how the vendor handles their data.
Expose a clear Privacy Policy link on the official homepage.
No terms link was found on the stated homepage. Buyers cannot see the conditions that apply to using the service.
Expose a clear Terms link on the official homepage.
fast-uri 3.1.7 (npm) has a known advisory: GHSA-hrr3-gc8f-f4qj (highest severity moderate). fast-uri vulnerable to inconsistent host case normalization via percent-encoded octets
Upgrade to a fixed version listed in the advisory and regenerate the lockfile.
smol-toml 1.8.0 (npm) has a known advisory: GHSA-r4xh-jqrq-34v2 (highest severity moderate). smol-toml: Quadratic-time parse() from parseKey rescanning to end of document on each key line
Upgrade to a fixed version listed in the advisory and regenerate the lockfile.
Each pillar starts at 100. Rule deductions reduce its score; weights determine the overall contribution.
No user reports yet.
Each report is one signed-in GitHub account saying what happened. We do not verify that the account ran the agent, and reports never change the automated score or grade. For a security problem, use the disclosure page instead of a public note.
Share the full report so its scope and limits travel with the score.
