Skip to content

Security: Rumeasiyan/askwell

Security

SECURITY.md

Security

Status

Askwell has no released version and no application code yet. There is nothing deployed to attack. This policy exists so it is in place before the first release rather than written after an incident.

Reporting a vulnerability

Use GitHub's private vulnerability reporting on this repository — the Report a vulnerability button under the Security tab. It is private to the maintainer and does not create a public issue.

Please do not open a public issue for a security problem.

Include what you can: what the issue is, how to reproduce it, what an attacker gains, and the version or commit. A proof of concept helps but is not required.

What to expect: a single maintainer, not a team. Acknowledgement within a week. No bounty programme. Credit in the release notes if you would like it.

What counts as a security issue here

Askwell runs entirely on one person's own machine, with no server and no accounts, so the usual web threat model mostly does not apply. What does:

Area Why it matters
Sandbox escape An imported database dump reaching outside its isolated database, Askwell's own data, or the network (C3). A dump is executable code and this containment is the only thing between it and the user's material
SQL validation bypass Anything that is not a single read query reaching the driver (C2). The user's real database is on the other side
Unexpected network egress Any outbound call in local mode, from any component or dependency (C1). This is the product's central promise and a breach of it is the most serious class of bug here
Prompt injection with effect Retrieved content causing real tool calls against the user's data (C7). Mitigations are documented as mitigations rather than solutions
Credential exposure Database credentials or the at-rest passphrase recoverable from disk, logs, traces or exports
Audit chain forgery Tampering that the hash chain fails to make evident (C6)

What does not count

  • Physical access to an unlocked machine. Askwell is a local application; someone at your keyboard has already won. The optional passphrase protects a stolen, powered-off laptop, nothing more.
  • The user's own choices — pointing Askwell at a model that behaves badly, or lowering the retrieval threshold. Both are permitted, both state their consequences.
  • Denial of service against yourself. A large import that fills your own disk is a usability bug, and worth reporting as one.

What we claim, and what we do not

The audit log is append-only and tamper-evident, not immutable. You own the machine and can always delete a file. The honest guarantee is that Askwell never rewrites history and that manual tampering is detectable. Anyone claiming more than that in documentation or marketing is making an error worth reporting.

There aren't any published security advisories