Current controls
Strict server-side workspace authorisation on every request; Argon2id password hashing; hashed session tokens; secure, httpOnly cookies; origin checks on state-changing requests; per-IP rate limits; Zod validation on all inputs; security headers (HSTS, CSP, X-Frame-Options, COOP, Referrer-Policy, Permissions-Policy); audit logging with secret redaction; field-level encryption for stored secrets; verified Stripe webhooks; least-privilege agent tools.
Prompt-injection defenses: user text, documents, websites, emails, imported agents, and tool output are treated as untrusted data — delimited content cannot change platform instructions, grant tools, read secrets, alter approvals, trigger billing, send, publish, or delete.
Files and generated code
File uploads are disabled in the current deployment because permanent private storage is not configured. The application does not fall back to public or ephemeral storage. Generated code must run in a separate disposable sandbox with no host mounts and network disabled unless a human approves a specific destination.
Security roadmap
Q3 2026 — Independent threat model and dependency/SBOM review.
Q4 2026 — External penetration test and managed rate limiting (edge WAF).
Q1 2027 — Managed secret manager migration, sandbox escape review, and quarterly restore drills.
Report a vulnerability
Email security@useoperatorhq.com. We acknowledge reports within 72 hours and keep you updated until resolution.
Disclosure policy: give us reasonable time to fix before public disclosure; do not access, modify, or delete other customers' data; do not send customer data, passwords, tokens, or active exploit details beyond what is needed to demonstrate the issue. We do not pursue legal action against good-faith security research that follows this policy. No PGP key is published at this stage — the mailbox is monitored by the founding team.