Do not file a public issue. Security reports should be emailed privately to security@awaithumans.dev.

If the vulnerability is under active exploitation or you need PGP-encrypted communication, note that in your first email and we'll coordinate a more secure channel.

We aim to:

• Acknowledge receipt within 2 business days.
• Provide an initial triage assessment within 5 business days.
• Ship a fix and a coordinated disclosure within 30 days for high and critical severity issues, faster when feasible.

Anything that affects the confidentiality, integrity, or availability of an awaithumans deployment — the SDK, server, dashboard, CLI, or bundled adapters / channels. A non-exhaustive list of the kinds of issues we want to hear about:

• Authentication or authorization bypass
• Session cookie / admin bearer token weaknesses
• SQL injection, command injection, path traversal
• Cross-site scripting, CSRF, clickjacking
• Slack signature verification weaknesses
• Secret exposure (encrypted columns, log leakage, response payloads)
• Timing side channels on auth paths
• Denial-of-service vectors against a single-tenant deployment
• Supply-chain issues in pinned dependencies
• Missing security headers or TLS misconfiguration defaults

• Self-inflicted misconfigurations (e.g. running without PAYLOAD_KEY in production, setting CORS_ORIGINS=* with credentialed requests, exposing the admin bearer token).
• Denial-of-service from an authenticated operator (operators are trusted by design — the dashboard admin surface is not a privilege escalation target).
• Rate-limit concerns on the login endpoint — v1 documents the absence of rate limiting; proper limiter lands in a post-launch hardening pass. A report pinning a specific amplification attack beyond online brute-force is in scope.
• Third-party services (Slack, Resend, Anthropic) — file with the vendor. We can relay if the issue is in how awaithumans calls into them.

When a fix ships, we credit the reporter in the release notes unless you prefer to stay anonymous. CVE assignment happens on request for any fix with a meaningful security impact.

• A response that takes the report seriously regardless of severity.
• A straight answer if we disagree with a classification, with our reasoning.
• No legal threats or takedowns for good-faith security research.
• Credit on disclosure if you want it.