AEGIS is pre-audit research software. It has not received an independent security review. Do not use it to protect people at real risk.

Known gaps are published in AEGIS_due_diligence.md §6 rather than withheld. Anyone evaluating AEGIS should read that section first. G3, G4, G5 and G7 are fixed with regression tests, and G6 (resource bounds) is partially fixed. The most serious currently-open items are delivery semantics / replay handling (G8), parser fuzzing (P1-4), and migration to audited native post-quantum crypto (G11/P1-5).

Report privately to: ideatrino@proton.me (A PGP key fingerprint can be added here if you use one.)

Please include a description, affected version or commit, reproduction steps, and your assessment of impact. Do not open a public issue for an exploitable finding.

• Acknowledgement within 3 business days.
• An initial assessment, with severity, within 10 business days.
• Regular updates until resolution.
• Credit in the release notes and advisory, unless you prefer otherwise.

We follow coordinated disclosure. Our default is to publish an advisory once a fix is available, targeting 90 days from report. We will move faster for actively exploited issues and will agree an extension with you if a fix proves genuinely difficult.

In scope: the protocol design, the aegis/ implementation, aegis_app.py, and aegis_chat.py.

Out of scope: endpoint compromise (malware, coerced unlock), denial of service against a local test node, and issues in third-party dependencies — please report those upstream, and tell us so we can pin or migrate.

kyber-py and dilithium-py are reference-quality pure-Python implementations that are not constant-time or side-channel hardened, and are slated for replacement with audited native libraries. Findings that depend on this are known (see gap G11 and checklist item P1-5) but we still want to hear about them.