How to report a vulnerability in the Plotwork self-hosted engine image, and what to expect. It is for anyone who finds a security problem in an image published from this repository.
| Question | Answer |
|---|---|
| Where do I report? | Use "Report a vulnerability" on the Security tab of this repository. That report is private. |
| Is there an email address? | No. The private report above is the only channel, so every report stays private until a fix is out. |
| Which versions get fixes? | The latest published version. |
| Should I open a public issue? | No. Public issues are for bugs that are not security problems. |
The image runs the diagram engine as one Node process, as user 65532, with a read-only filesystem. It serves MCP at /mcp and the REST drawing routes under /v1, behind a bearer token, and /healthz. It makes no outbound request, sends no telemetry, stores nothing, and logs no diagram text. A report is most useful when it shows one of the following:
- the server reaching the network, writing to disk, or starting another process
- a request that gets past the token, or past the origin check from a browser
- a tool argument or request body that makes the server misbehave, or exhaust memory or CPU past its limits
- diagram text in a log line
- a secret or a path of a build machine inside the image
- an image whose files do not match
checksums.txt, or whose signature or attestation does not verify
Give the image version and digest (docker inspect --format '{{index .RepoDigests 0}}' <image>), how you run it, the smallest request that shows the problem, and what you expected instead.
The maintainers acknowledge a report, confirm the problem, and publish a fixed version. A published image is never changed or removed. A fix is a new version, and the release notes of an affected version point to it.
| Link | Why |
|---|---|
| README | What the image is, how to run it, and how to verify it |
| Self-hosting guide | Every setting, Compose, Kubernetes, and connecting agents |