Quaestor exists to make security claims defensible. That standard applies to the platform itself: what a model sees, where your code goes, and how every verdict is signed.
Each gate decision (A–H) is Ed25519-signed with a per-tenant key over a canonical payload hash. A sealed manifest binds the run to the exact commit and the exact sanitised bytes. Nothing in a report is asserted — it is signed, and re-verifiable offline.
gate · payload sha256 · signer · Ed25519
Before any model reads a line, Gate 0 clones and sanitises the repository — stripping secrets and PII and recording every redaction (file, line, detector) into the signed manifest. Discovery, validation, and proof reason only over the sanitised tree.
clone → strip → seal → (then) analyse
On-Prem runs single-tenant inside your boundary. SaaS enforces hard tenant isolation server-side on every query — the tenant is derived from the session, never a client-supplied header, and a foreign identifier returns 404, never a disclosure. Passwords are PBKDF2-HMAC-SHA256; sessions are HttpOnly · Secure · SameSite.
On-Prem has no outbound dependency to run the pipeline. LLM inference is your choice: an on-host model, a metered API under your own contract, or your existing subscription via a host-side agent — your credentials never enter a container, and only the sanitised prompt ever leaves the sandbox.
A finding that can't be proven is refused, not shipped. Disproven cases stay visible — rigor is a feature, not an embarrassment. There are zero rubber-stamped verdicts: a case cannot be reported without passing the signed gates that earn it.
Exports are content-addressed and reproducible: the same evidence renders the same signed artefact. Every mutation is written to an immutable audit log with actor and role. A board report never contains a claim the pipeline didn't earn.
We welcome coordinated disclosure. Email security@quaestor.dev with details and reproduction steps; we acknowledge within two business days and will keep you updated through remediation. Please don't test against tenants you don't own.
no bounty program yet · good-faith research credited
See a signed case end to end, then run your own repository through the same pipeline.