Sailfin is a systems language whose core promise is compile-time capability enforcement — every function declares the effects it uses (![io, net, clock, ...]) and the compiler rejects code that exceeds its declaration. A vulnerability in the compiler, runtime, or toolchain can undermine that promise (for example, by letting code perform an effect it never declared, or by corrupting memory in the native runtime). We take such reports seriously.

Sailfin is pre-1.0 and ships on an alpha channel. Security fixes land on main and are delivered through the most recent release only — we do not backport to older prereleases.

If you are on an older build, upgrade to the latest release (or main) before reporting — the issue may already be fixed.

Please do not open a public issue, pull request, or discussion for a security vulnerability. Public disclosure before a fix is available puts every user at risk.

Report privately through GitHub Security Advisories:

1. Go to https://github.com/SailfinIO/sailfin/security/advisories/new (or the repository's Security → Advisories → Report a vulnerability button).
2. Describe the issue with enough detail for us to reproduce it.

A good report includes:

• Affected component — compiler pass (lexer/parser/typecheck/effect checker/emitter/LLVM lowering), native runtime (runtime/native, runtime/sfn), the build/self-host driver, or a released binary.
• Version / commit — the release tag or git commit, and your platform (e.g. linux-x86_64, macos-arm64).
• Reproduction — a minimal .sfn snippet or command sequence that triggers the issue, plus expected vs. actual behavior.
• Impact — what an attacker gains (capability/effect bypass, memory corruption, code execution, denial of service, supply-chain compromise).

• Acknowledgement within 5 business days.
• An initial assessment (severity, affected versions, whether we can reproduce) within 10 business days.
• Coordinated disclosure: we'll work with you on a fix and a disclosure timeline. We aim to ship a fix within 90 days of triage, sooner for actively exploited issues.
• Credit in the published advisory, unless you prefer to remain anonymous.

If you do not receive an acknowledgement within the window above, please follow up on the advisory thread.

Security-relevant areas for this repository include, but are not limited to:

• Effect / capability bypass — code that performs an undeclared effect (![io], ![net], ![model], ![clock], …) without a diagnostic, or a way to defeat capsule-manifest capability enforcement. This is the project's central guarantee; bypasses are high severity. Note: effect enforcement is fully realized on Linux x86_64 and partial on macOS arm64 (tracked in public issues) — known gaps documented there are not new vulnerabilities, but a novel bypass on a platform claimed to enforce is in scope.
• Native runtime memory safety — heap corruption, use-after-free, out-of-bounds access, or ownership/aliasing-analysis soundness holes in runtime/native/ and runtime/sfn/.
• Compiler miscompilation with a security impact — generating code that silently violates the source program's declared safety properties.
• Supply chain — the seed/self-host trust chain, release artifacts, the capsule dependency resolver, or build tooling that could let a malicious dependency or seed compromise a build.

• Bugs with no security impact — file a normal issue instead.
• Features documented as not yet enforced (e.g. PII<T> / Secret<T> runtime enforcement, AI/model blocks) — these are tracked publicly and are known-incomplete, not vulnerabilities. See docs/status.md and CLAUDE.md for what currently ships vs. what is parsed-but-not-enforced.
• Vulnerabilities in third-party tools we invoke (clang, lld/mold, LLVM) — report those upstream; if our usage makes them exploitable in a way they otherwise wouldn't be, that part is in scope.

We will not pursue or support legal action against researchers who, in good faith, follow this policy: who avoid privacy violations, data destruction, and service disruption, and who give us a reasonable opportunity to remediate before any public disclosure.

Thank you for helping keep Sailfin and its users safe.