Modgud is an Identity Provider — a compromise here has outsized blast radius for downstream applications. We take security reports seriously and want to make it easy for researchers to do the right thing.

In scope:

• The Modgud IdP itself — backend (src/dotnet/Modgud.Api, Modgud.Authentication, Modgud.Authorization, Modgud.Domain, Modgud.Infrastructure) and admin SPA (src/frontend-vue/).
• The Modgud.Client.AspNetCore NuGet package that downstream apps use to validate Modgud-issued tokens.
• The official Docker image (ghcr.io/cocoar-dev/modgud:*).
• The default configuration shipped in src/dotnet/Modgud.Api/data/configuration.json.

Out of scope:

• Third-party dependencies — please report those to their upstream (Marten, Wolverine, OpenIddict, Fido2NetLib, Vue, etc.). If a vulnerability in a dep affects Modgud, that's in scope here too and we'll coordinate.
• Build-time and test-only dependencies. Vulnerabilities in packages that exist solely in our build pipeline or test harness (e.g. VitePress build tools, testcontainers and its transitive graph including protobufjs/grpc-js, @vue/test-utils, ESLint plugins) are not treated as security issues — the vulnerable code paths never run in the shipped runtime. We track them via the weekly Dependabot sweep on the production manifests and patch on a best-effort cadence. See .github/dependabot.yml for the exact scope.
• Deployments not run by COCOAR e.U. — if you find an issue with a third-party Modgud instance, please contact that operator first.
• Recovery scenarios involving server access — the recover CLI intentionally bypasses every UI auth check; treat container/file access as already-compromised.
• Self-XSS or attacks requiring already-authenticated admin (realm:admin) cooperation.
• Issues in legacy-final (the pre-rebuild snapshot tag) — preserved only for historical reference, not supported.

Preferred channel: GitHub Private Vulnerability Reporting.

Open a private advisory on cocoar-dev/modgud — the form lets you describe the issue, attach a PoC, and propose fixes in a private space only the maintainer can see.

Alternative: email bwi@cocoar.dev. PGP welcome but not required; ask for our public key in the first message if you want to use it.

• A minimal reproduction (curl command, payload, code snippet, browser steps — whichever fits)
• Affected versions or commit SHAs
• Impact assessment (what does an attacker get?)
• Suggested mitigation if you have one — extra credit, not required
• How you'd like to be credited (real name, handle, or anonymous)

Response is best-effort — we triage as fast as we can but don't run a 24/7 SOC. Please don't take silence as dismissal: if you haven't heard back within the acknowledgement window, send a one-line ping.

Good-faith security research conducted under this policy will not be pursued legally. We will not threaten, sue, or report a researcher who:

• Tests only their own Modgud installation (or one they have explicit permission to test)
• Avoids accessing, modifying, or destroying other users' data
• Avoids degrading service for other users
• Reports findings privately through the channel above before public disclosure
• Gives a reasonable timeframe for a fix before publishing

If unsure whether an action would qualify, ask first — privacy or denial-of-service tests against COCOAR-operated infrastructure require prior written agreement.

We prefer coordinated disclosure: report → triage → fix → release → public advisory.

• A GitHub Security Advisory will be published when the fix lands.
• A CVE will be requested via GitHub's CVE programme for any vulnerability with CVSS ≥ 4.0, unless the reporter prefers otherwise.
• Reporters are credited in the advisory by their chosen attribution.
• We aim to publish the advisory within 14 days of the fix landing.
• If a fix takes substantially longer than the targets above, we'll agree a coordinated public-disclosure date with the reporter rather than letting it sit indefinitely.

A SAST / CodeQL / dependency-audit sweep closed 31 of 33 findings before this policy went into effect. PII masking, CSP, anti-CSRF middleware, path-traversal guards, and rate-limiting on credential endpoints are documented in the public security model — see docs/concepts/authentication.md for the relevant cookies + middleware, and the dev-only security hardening tracker for the per-finding history.

This doesn't mean we're free of bugs — it means we know where to look when one is reported.