This is a rolling, single-main-branch deployment template; only the latest release on main receives security fixes. If you are running an older revision, upgrade to the newest version to pick them up.

Please do not file a public GitHub issue for security problems. Instead, report it privately through GitHub's private vulnerability reporting: open the repository's Security tab and click "Report a vulnerability". This creates a private advisory visible only to you and the maintainers. (If that option is not visible, open a minimal public issue asking for a private channel, without disclosing any details.)

Please include:

• A description of the issue
• Steps to reproduce, including a minimal example if possible
• The affected version (commit SHA if known)
• Any suggestions for remediation

Reports will be acknowledged within a reasonable window (target: 5 business days). Time-sensitive issues should say so. Please allow up to 90 days for a coordinated fix before public disclosure.

This project is a deployment template with two auth planes (AUTH_MODE):

• access (default, private self-host): Cloudflare Access protects the worker URL; the worker trusts Cf-Access-Authenticated-User-Email for per-user scoping.
• public (hosted play.skyphusion.org): first-party username/password accounts with opaque session cookies; no CF Access; per-user BYOK only (worker gateway secrets ignored). BYOK is account-scoped (v1.1.0): calls go to the user's own Cloudflare account ID over HTTP, because the AI binding can only reach the worker's account. Credentials without an account ID are refused (412), never resolved on the host binding.

In both modes:

• D1 history is scoped per user identity (Access email or opaque account id)
• R2 objects carry customMetadata.user_email for ownership checks

In-scope vulnerabilities include:

• Bypasses of the per-user history scoping
• R2 object access by a user other than the owner
• Authentication or authorization issues in the artifact-serving path
• SQL injection or other injection issues
• Cross-site scripting or content injection via stored attachment metadata
• Logic errors that leak data across users

Out-of-scope:

• Issues that require already-compromised Cloudflare Access credentials
• Denial-of-service via legitimate-but-expensive Workers AI calls (this is a billing concern; rate-limit at the Gateway if needed)
• Issues in upstream Cloudflare services themselves