Reporting a vulnerability
If you have found a security problem in MiraServe, please report it privately so it can be fixed before it is disclosed. This page explains where to send it, what to include, and what is in scope.
How to report
MiraServe is in private testing and has not been released, so there is no public version to report against yet. A private reporting address will be published on this page with the first release.
When it opens, reports will be visible only to you and the project maintainers. Do not post details of a security problem publicly until a fix is available.
MiraServe is alpha software (0.1.0-alpha.0) with no tagged release. A formal response-time commitment and disclosure policy have not been published yet.
What to include
- The commit hash or version you tested (
miraserve --version), and the operating system. - The configuration involved, reduced to the smallest file that shows the problem, with secrets removed.
- Exact steps or a script to reproduce it, and the requests sent.
- What happened and what you expected.
- Your assessment of the impact: what an attacker gains, and what access they need first.
- Whether you would like to be credited, and under what name.
In scope
Anything in the MiraServe source code once it is published, in particular weaknesses against the threat model:
- Crashes, hangs or memory exhaustion triggered by network input or a configuration file.
- Request smuggling, or HTTP/2, HTTP/3 and QUIC denial-of-service limits that do not hold.
- Reading files outside a configured document root.
- Exposure of TLS private keys to the worker process in split-daemon mode, or any way to obtain signatures from
miraserve keydwithout being the permitted user. - Publishing or rolling back a site without valid credentials, bypassing signature verification, or getting content past a pre-flight check it should fail.
- Calling a tool the MCP gateway should refuse, or bypassing its token checks.
- A WASM plugin escaping its limits or reaching files, network or environment.
- Secrets written to logs despite redaction being on.
- A configuration that
miraserve checkaccepts although it violates a rule the project documents as enforced. - Weaknesses in the generated service units or packaging definitions.
Out of scope
- The non-goals listed in the threat model: bugs in your own backend applications, attacks that need root on the host or a compromised kernel, and network floods that saturate the link.
- Problems that only appear after you turn off a protection that
checkwarns about, unless the warning is missing. - Vulnerabilities in third-party dependencies that do not affect MiraServe. If one does, please report it here as well as upstream.