Security Policy

Status: pre-1.0, recently open-sourced

mdd is pre-1.0 and was published as open source in 2026. Please read this before you deploy it anywhere that matters:

• There has been no independent security review of this codebase.
• mdd is written almost entirely by AI agents under human review. The review is real, but it is not the same thing as an audit.
• Under 0.x semantic versioning, breaking changes land between minor versions. Pin a version if you need stability.

None of that means we are relaxed about security reports. It means you should calibrate your trust accordingly, and that we would rather hear about a problem than not.

Supported versions

Version	Supported
latest 0.x	yes
any earlier 0.x	no

Security fixes land on main and ship in the next release of the current minor. While mdd is pre-1.0 there are no security backports to older minor versions. This will be revisited at 1.0.

Reporting a vulnerability

Please report privately. Do not open a public GitHub issue, and do not discuss the vulnerability publicly until a fix is available.

Preferred: email abuse@schubergphilis.com with [mdd] in the subject. This is Schuberg Philis’s standing responsible-disclosure channel. Encrypted mail is preferred; the PGP public key is at https://keybase.io/schubergphilis, and an encrypted chat option is linked from the page below. The company’s full responsible disclosure policy — including credit, the hall of fame, and the bounty — is at https://schubergphilis.com/security and takes precedence over this document on anything it covers.

On GitHub: if you would rather report through GitHub, private vulnerability reporting is enabled on this repository — use the “Report a vulnerability” button under the Security tab.

Please do not send sensitive information over unencrypted channels or social media.

Please include:

• What the issue is and what an attacker gets out of it.
• Steps to reproduce, or a proof of concept.
• The mdd version (mdd --version) and platform.
• Any workaround you have already found.
• Whether you want public credit or prefer to stay anonymous.

Do not include real customer data, live credentials, or content from a private Confluence space or SharePoint site in a report. A redacted reproduction is more useful to us than a real one.

Rules of engagement

If you think you have found a vulnerability, we would very much appreciate it if you:

• Don’t exploit your finding.
• Share the information with just us, not other parties.
• Give us time to assess the situation and respond within a reasonable timeframe.

What to expect

mdd is maintained on a best-effort basis by a small team. We aim to:

1. Acknowledge your report within 5 business days.
2. Confirm or close it, with a written explanation either way, within 15 business days.
3. For confirmed issues, agree a fix and disclosure timeline with you. Severe issues in the credential and push paths get priority; expect a release rather than a patch to an older version (see Supported versions).
4. Credit you in the release notes and the advisory if you have told us you want public credit. By default you remain anonymous.

These are targets, not contractual commitments, and they apply to this open source project. Reports about Schuberg Philis production systems fall under https://schubergphilis.com/security instead.

Credit and bounty

By default you remain anonymous. If you would rather be named, say so and tell us the name or alias to use; we will credit you in the GitHub advisory and the release notes.

Reports sent to abuse@schubergphilis.com are handled under Schuberg Philis’s responsible disclosure policy, which offers reporters a token of appreciation and maintains a public hall of fame. See https://schubergphilis.com/security for the current options and what it covers. mdd runs no separate bounty of its own.

Scope

In scope: the mdd package and CLI in this repository, and documentation that would lead a careful operator into an insecure configuration.

Out of scope:

• The services mdd talks to — Confluence Cloud, SharePoint / OneDrive / Microsoft Graph, 1Password, GitLab, GitHub, and any OpenAI-compatible AI gateway. Report those to the respective vendor.
• External tools — Quarto, librsvg, ripgrep, Microsoft Office, op, git. Report upstream.
• Third-party Python dependencies. Report upstream. Do tell us if the way mdd calls a dependency turns a non-issue into an exploitable one, or if we are pinning something known-vulnerable — pip-audit runs in CI, but it is not infallible.
• Operator misconfiguration, including an empty or wrong confidentiality blacklist, an over-broad Confluence or Graph token, a world-readable config directory, or pointing origin at a remote you did not mean to publish to.
• AI output quality. The mdd ai commands call a third-party gateway; output is non-deterministic and may be wrong. That is a known property, not a vulnerability.
• Breaking changes between 0.x minors. Documented above.
• Findings from a scanner with no demonstrated impact on mdd.

Good faith

If you follow this policy in good faith, we will treat your report as a contribution: we will work through it with you, we will tell you what we decided and why, and we will not ask you to stay quiet indefinitely. In return, please don’t access, modify, or retain data that isn’t yours, and don’t degrade any service while investigating.