Report a security vulnerability
A single entry point for every MoX project (moxsh-terminal, moxsh-proot, moxwebgpu, agent-core, moxsh-plugins, moxsh-suite, mox-site, mox-id). No need to dig through six SECURITY.md files, and no GitHub account required — this form supports fully anonymous reports.
Submitted over HTTPS and written only to a server-side database, the report is readable in cleartext by administrators in the admin console after authentication — no encryption for you, no decryption for us, and nothing exposed to the browser. For a directly exploitable 0day, describe impact and reproduction steps only; send full exploit code separately after we confirm a channel.
Code-of-conduct complaints are not vulnerabilities. Use the separate conduct complaint form; the two queues are deliberately kept apart.
Submission channels
| Anonymous form below | Recommended. No account and no encryption needed. You immediately receive a one-time ticket to check progress. IP rate limiting and human verification block bulk abuse. |
|---|---|
| GitHub private reporting | Security → Report a vulnerability on each public repo. Requires a GitHub account, but feeds directly into a GitHub Security Advisory and can be linked to a CVE. |
| Contact email | Revealed after sign-in on the homepage Contact section. For general correspondence only — prefer this form for vulnerability details so they are not left in a mailbox. |
Why no encryption is required
We deliberately do not use PGP: encrypted reports force the maintainer to decrypt every submission before reading it in the admin console, which slows triage and makes reports easy to miss. Instead we constrain both transport and access:
- Transport: HTTPS end to end; an edge function verifies human-check and rate limits before forwarding, and strips credentials at the edge.
- Storage: the body is written only to a server-side database with no public read endpoint. The admin console requires an authenticated administrator session, and every action is audited.
- Receipt: submission returns only a one-time ticket (stored as a hash); no public page echoes the body or contact details.
If your organisation mandates end-to-end encryption, use GitHub private vulnerability reporting or contact us by email to agree on a separate channel.
Coordinated disclosure
| Acknowledgement | Usually within 72 hours; holidays and travel may delay it. This is a target, not a hard SLA. |
|---|---|
| Fix timeline | Usually an initial assessment and plan within 3 days of acknowledgement; high severity first. |
| Embargo | Default 30 days from the release containing the fix. Neither side publishes details during this window; ask if you need an extension. |
| 90-day fallback | If no fix and no reasonable explanation within 90 days of acknowledgement, you may publish without objection from us. |
| CVE | The maintainer decides whether to request one; reporters can be credited as discoverers. We do not accept paid disclosure conditioned on a CVE. |
| Credit | Named in the CHANGELOG and the hall of fame below on release (anonymous or pseudonymous on request), including low-severity and duplicate reports unless declined. |
| Notification | Announced via GitHub Security Advisories, Release Notes and the status page; CLI and App update checks surface security upgrades where available. |
Supported versions
These projects move fast: only the latest release on each repo's default branch is maintained. Older versions receive no separate patches — upgrading is the fix. Version numbers live in each repo's Releases, package.json or Cargo.toml; we deliberately do not freeze them on this page.
- moxsh-terminal (Android app, GPL-3.0)
- moxsh-proot (Rust engine, GPL-3.0)
- moxwebgpu / agent-core (npm packages, MIT)
- moxsh-plugins / moxsh-suite (spec and umbrella repos, MIT)
- mox-site / mox-id (website and identity backend, continuously deployed, unversioned)
Threat model: what counts
- moxsh-proot: container escape, bind-mount traversal into the host filesystem,
/proctraversal leaking host paths, forged entitlement signatures or paid-gate bypass, recipe command injection, backup/restore destroying container data. - moxsh-terminal: plugin sandbox escape, undeclared permissions being exercised, malicious
.moxpackages, local credential or session-token disclosure, clipboard and terminal escape-sequence injection. - mox-id / mox-site: cross-account data access, session fixation / CSRF, open redirects, OAuth state bypass, rate-limit bypass, privacy fields being smuggled into storage.
- Out of scope: an attacker who already fully controls the local device, pure social engineering, missing features, upstream issues with no feasible mitigation, and scanner output with no demonstrable impact.
Anonymous report form
Check progress
Hall of fame
No publicly credited reports yet. After a fix ships, reporters are listed here as they prefer (anonymous or pseudonymous).
Historical advisories
See each repo's GitHub Security Advisories and Releases; this page does not duplicate a list that would go stale.