Security
Reporting a vulnerability
Report a vulnerability privately, by either of these:
- GitHub private vulnerability reporting. Open
a private report on the kubemoot repository,
or use Report a vulnerability on the Security tab of the affected repository
(
kubemoot,kmctl,vscode-crewforge,crews, orkubemoot-docs). The report is sent over HTTPS and only you and the maintainers can see it. - Email security@kubemoot.org. The project publishes no OpenPGP key, so email is not encrypted end to end; when the details must stay confidential in transit, use GitHub’s private reporting.
The security address is only for vulnerability reports; send every other question to moot@kubemoot.org or GitHub Discussions. Do not open a public issue, Discussion, or pull request for a security report.
Include what is affected (repository and version), the steps to reproduce, and what an attacker gains. The project acknowledges reports within a reasonable period and coordinates disclosure with the reporter before any public announcement. With a single maintainer, that is best effort, not a guaranteed response time.
Supported versions
Every merge to main that changes a component builds a release candidate, and a maintainer publishes it as a
versioned release (see Releases). Kubemoot is a young project and does not backport fixes.
A security fix lands on main and ships in the next published release of the affected component;
to receive it, move to that release.
How the code is checked
These checks run on the public repositories without anyone starting them:
- CodeQL code scanning, for every language in each repository, on every pull
request, every push to
main, and weekly. - Dependabot security updates and version updates for each repository’s dependencies and GitHub Actions.
- Secret scanning with push protection, which blocks a push that contains a known credential format.
- OpenSSF Scorecard, on every push to
mainand weekly. The results are public, for example the Scorecard for kubemoot. - Fuzzing. Every Go native fuzz target in the operator runs on each operator change,
in pull requests and on
main. - SonarQube quality gate on every push to
maininkubemoot,kmctl,vscode-crewforge, andcrews; forkubemootit also imports Trivy’s findings.
Workflows reference GitHub Actions by commit SHA, and the default workflow token is read-only; a job that needs more asks for it.
Every container image and Helm chart that Publish Release publishes to GHCR carries
a keyless Sigstore signature and SLSA build provenance
(Verify images and charts). The kmctl and
CrewForge releases carry the same for their release assets, attached to the GitHub
Release.
For the full picture across change control, releases, supply chain, and settings, see How Kubemoot Is Built and Secured.
Security model and known limitations
Kubemoot is a v1alpha1 platform to evaluate and shape, not one to install where it
holds data or credentials you care about without reading its security model first. The
limitations are listed as they stand so you can decide whether they matter for your
deployment. The full text is the “Security model and known limitations” section of
SECURITY.md in the
kubemoot repository. It covers:
- In-cluster endpoints, including the NATS message bus, that accept requests without authentication, and the fact that namespaces separate names but are not a security boundary today.
- Creating an
Agent,MCPServer, orRAGSourceis equivalent to creating pods. - The RBAC of the optional internal MCP servers and of the reference crews.
- Agents act on text they read, so prompt injection through a tool, document, or web page can steer a tool-calling agent.
- The code sandbox depends on a CNI that enforces NetworkPolicy.
- API stability and platforms:
kubemoot.ai/v1alpha1can change between releases, and release images areamd64only.
Admission webhooks require cert-manager; the quickstart runs with webhooks off for that reason.