kint holds the key to someone's memory, so a flaw in it is a flaw in their custody. If you find one, please report it privately.
Open a private advisory at
github.com/s0nderlabs/kint/security/advisories/new.
Include the version (kint --help prints it in the package metadata; the tag is in
pyproject.toml), what an attacker needs, what they gain, and a way to reproduce it. Do not open a
public issue for anything that can expose a vault key, a session key or memory contents.
src/kint/(key derivation, the envelope, push, pull, restore, verify, the MCP server, the CLI and the loopback authorize page).contracts/src/EpochAnchor.sol, deployed on Base mainnet at0xa22E03f7a4145Bf4909a83595C90a38E14d79600. It has no admin and no upgrade path, so a fix is a new deployment and a new chain of epochs.js/(@s0nderlabs/kint-core, the browser decrypt path).
Sibyl Memory itself is out of scope here; report issues in it to Sibyl Labs.
These are documented properties, not vulnerabilities: a phished derive signature is a permanent
key; epochs sealed before a kint rekey stay readable to whoever held the old key; the row count
and the size bucket of every epoch are public; reading the ciphertext back needs a node that keeps
its transaction index. The full threat model, with who can do what, is the
Limits and threat model chapter.